Friedberg

The prediction’s deadline is “by the end of 2026”, so as of the current date (30 November 2025) there is still over a year left for it to play out.

Public information shows that:

• Huawei already designs and sells Ascend AI accelerators (e.g., 910B/910C) produced by SMIC on a 7 nm process. These chips are used in large clusters, are priced below Nvidia’s top GPUs in China, and are positioned as competitive—though generally behind Nvidia’s latest H100/B200 in absolute performance and efficiency—for certain AI workloads.(tomshardware.com)
• Huawei and SMIC are constrained to older nodes (around 7 nm) because they lack access to EUV lithography; reporting indicates their next Ascend generations are still planned on 7 nm-class tech through at least 2026, with ongoing yield and capacity challenges.(fortune.com)
• Chinese equipment maker SiCarrier has publicly described DUV multi‑patterning techniques that could enable 5 nm‑class production without EUV, and says its tools are used by SMIC and in collaboration with Huawei—but these techniques are openly discussed in industry reporting, not secret or undisclosed.(reuters.com)
• Huawei is reportedly involved in a “secret” network of fabs and proxy firms to expand domestic chip manufacturing under U.S. sanctions, but articles describing this network do not provide evidence of a novel, non‑public lithography technology already deployed in volume production.(bloomberg.com)

So, while parts of the prediction are partially aligned with current reality (Huawei producing domestic AI accelerators at significant scale that are somewhat competitive and cheaper for some markets), the specific claim about a not publicly discussed Chinese lithography technology being deployed in new mainland fabs is not supported by current public evidence, and the time window for the overall prediction has not yet closed.

Because the deadline (end of 2026) lies in the future and there is still ample time for events to change—either toward confirming or refuting the claim—the outcome cannot yet be definitively judged. Hence the result is inconclusive (too early).

It is too early to evaluate this prediction.

The claim is about the end of the 21st century (by ~2100). As of today (2025), electron-based semiconductor computing still overwhelmingly dominates all major computing workloads. Photonic and quantum–photonic technologies exist but are at early research or niche-deployment stages:

• Silicon photonics is being used mainly for high-speed data center interconnects and specialized optical networking, not as a general replacement for CMOS logic in CPUs/GPUs.
• Photonic computing prototypes (e.g., optical matrix multipliers and accelerators for AI) are still experimental or very limited in commercial deployment, far from handling “most major computing workloads.”
• Quantum and quantum–photonic systems are early-stage: current devices have small qubit counts, high error rates, and are used for research and very limited proof-of-concept applications, not broad production workloads.

Because the prediction is explicitly about the technological landscape by ~2100 and we are only in 2025, there is no way to determine yet whether electron-based computing will be “largely supplanted” for most workloads. The correct status is therefore inconclusive (too early to tell).

As of November 30, 2025, the proposed one‑time 5% wealth tax on California billionaires exists only as a proposed initiative, commonly called the 2026 Billionaire Tax Act or California Billionaire Tax Act. Organizers filed it with the California Attorney General on October 21–22, 2025 and are just beginning to gather signatures to qualify it for the November 2026 ballot.(seiu-uhw.org)

Current reporting and campaign materials describe the measure’s substance and signature requirements but do not indicate that:

1. Voters have yet voted on it (it has not appeared on any ballot), or
2. Any court has ruled on its constitutionality or blocked it from taking effect (it has not even been enacted).(axios.com)

Some policy analysis pieces explicitly say that the tax would likely face constitutional challenges in the future, but they describe anticipated litigation, not any decided case.(californiapolicycenter.org)

Because the key contingencies in Friedberg’s prediction—voters approving the amendment and courts then finding it unconstitutional so it never goes into effect—have not yet occurred or been tested, there is not enough information to declare the prediction right or wrong at this time.

As of November 30, 2025, the SEIU-backed “2026 Billionaire Tax Act” has only recently been filed with the California Attorney General and is still in the early procedural phase. It is undergoing the 30‑day public review, after which proponents will have up to 180 days to gather roughly 875,000 signatures to qualify for the November 2026 ballot.(insidesalt.com)

Coverage of the initiative consistently frames SEIU‑UHW’s stated purpose as replacing Medicaid/Medi‑Cal funding lost due to recent federal cuts and averting a health‑care crisis, with revenue directed 90% to Medi‑Cal and 10% to K‑12 education.(apnews.com) These reports describe messaging around protecting health care and education, not an explicit strategy of baiting opponents so they can be politically “destroyed.”

Because the next election cycle in which this measure would be on the ballot is November 2026, that campaign has not yet begun in earnest. There is not yet evidence of systematic SEIU or allied campaigns using the measure to target and politically damage high‑profile figures who publicly denounce it (for example, no documented attack‑ad campaigns or coordinated hit pieces tied to specific denouncers as of now).

Given the short time between the podcast (October 24, 2025) and today, and the fact that the 2026 election cycle is still ahead, it is too early to determine whether Friedberg’s prediction about the initiative’s primary use as political bait will ultimately prove correct.

California’s Proposition 55 (the 2016 California Children’s Education and Health Care Protection Act) already extends the higher income‑tax rates on high earners through 2030; any further extension beyond 2030 requires a new measure. (calbudgetcenter.org)

As of November 30, 2025:

• The California Teachers Association and allied groups have filed and are campaigning for a new initiative, The California Children’s Education and Health Care Protection Act of 2026, intended to permanently extend the Prop 55 high‑earner tax and place it on the November 2026 ballot. (cta.org)
• The California Secretary of State lists this measure (initiative no. 25‑0016) as a proposal pending review by the Attorney General, meaning it has not even reached the signature‑gathering completion stage, let alone appeared on a ballot or been approved by voters. (sos.ca.gov)

Because no law or ballot measure has yet been enacted to extend the Prop 55 tax rates beyond their current 2030 sunset, and the decisive 2026 vote has not occurred, we cannot say whether Friedberg’s prediction that “they’re going to get that passed” is right or wrong at this time.

As of November 30, 2025, only about six weeks have passed since the prediction (made on October 17, 2025) and essentially none of the 10‑year forecast horizon (“the decade ahead”) has played out. The claim is specifically about how US labor-market dynamics will unfold over the entire coming decade for workers in occupations threatened by AI and automation (e.g., drivers): that, in general, a recruiting cycle into new, higher‑paying roles will precede large‑scale elimination of the old roles.

Current empirical evidence and forecasting work mostly emphasize that AI adoption and its labor‑market effects unfold over many years, concentrated in the 2030s, not in the very short run:

• McKinsey and the World Economic Forum estimate that large portions of work could be automated, but suggest that the bulk of these effects will materialize between about 2030 and 2060, with a midpoint around 2045, implying a long diffusion and adjustment process rather than immediate mass displacement. (mckinsey.com)
• The Penn Wharton Budget Model finds that, as of 2025, AI’s impact on productivity and employment is still small; employment has only modestly fallen in the tiny share of jobs that are almost fully automatable, and slowed in highly exposed occupations, indicating early, mixed signals rather than a settled pattern of how transitions will work. (budgetmodel.wharton.upenn.edu)
• A Yale/Brookings analysis of the US labor market concludes that, since the rise of tools like ChatGPT in late 2022, there has not yet been significant, clearly attributable AI‑driven disruption; observed changes are largely consistent with pre‑existing trends, underscoring that any major reallocation of labor is still in its early stages. (theguardian.com)
• IMF work on AI and jobs stresses that around 60% of jobs in advanced economies may eventually be affected, but frames this as a forward‑looking risk over the next decade(s), highlighting the need for policies to manage worker transitions rather than documenting outcomes that have already occurred. (thenationalnews.com)

These sources collectively indicate that we are at the very beginning of the adoption curve and that large‑scale job reconfiguration due to AI is expected mainly later in the 2020s and 2030s. It is far too early to determine whether, on average across threatened occupations, US employers will in fact recruit workers into new, better‑paid roles before old roles are eliminated at scale. Some early data even show stagnation or slight declines in highly AI‑exposed jobs, but this is neither large‑scale nor clearly accompanied (or not) by the kind of systematic, prior recruiting cycle Friedberg describes. (budgetmodel.wharton.upenn.edu)

Because the prediction is explicitly about a full decade-long process and the relevant period has essentially not yet occurred, the correctness of the claim cannot be evaluated at this time.

As of November 30, 2025, it is far too early to definitively judge a 5–10 year prediction about U.S. power demand driven by AI.

What we can say based on current evidence:

• Multiple analyses (e.g., from the International Energy Agency and major U.S. utilities) project that rapidly growing data center and AI workloads will significantly increase electricity demand through the late 2020s and early 2030s, with some U.S. regions revising long‑term load forecasts sharply upward and explicitly citing AI/data centers as a key driver.
• However, these are projections covering much of the 2025–2035 period; they do not yet constitute retrospective evidence that a “very large increase” has actually occurred over the full window the prediction refers to.
• Only ~2 months have elapsed since the October 3, 2025 podcast release, and we are only at the very beginning of the stated 5–10 year horizon (2025–2035). Even if early signs and forecasts strongly support the prediction’s plausibility, it cannot yet be confirmed or falsified in outcome terms.

Because the time window has not come close to ending and the claim is explicitly about what will happen over the next 5 or 10 years, the correct status today is “inconclusive (too early)”, not “right” or “wrong.”

The prediction concerns (1) the mix of new U.S. electricity generation capacity built to serve AI demand over the next five years (roughly 2025–2030) and (2) the mix over the subsequent 5–10 years (roughly 2030–2035). As of the current date (2025‑11‑30), less than a year of the first five‑year window has elapsed, and none of the 2030–2035 window has begun. It is therefore too early to determine whether AI‑driven capacity additions through ~2030 will be predominantly natural‑gas‑fired, or whether nuclear will become a major component in the 2030–2035 period. Early policy moves or project announcements cannot conclusively establish the realized generation mix across those full future windows, so the prediction cannot yet be judged.

As of November 30, 2025, the preconditions Friedberg described are largely in motion, but his prediction is explicitly about a multi‑year restructuring, so it’s too early to know if it will “rewrite the entire media landscape.”

• Paramount/Skydance consolidation: Skydance’s takeover of Paramount Global closed on August 7, 2025, creating “Paramount Skydance” under David Ellison’s control, with significant cost‑cutting and strategic reshaping already under way. (en.wikipedia.org)
• Potential Warner Bros. Discovery acquisition: Paramount Skydance has made and is preparing further bids for Warner Bros. Discovery, and WBD is running an auction process (including other bidders like Comcast and Netflix). WBD has already rejected at least one Ellison‑backed offer, and as of late November 2025 no sale to Paramount Skydance has closed; WBD is still soliciting improved bids and weighing options including a breakup. (cnbc.com) This means the “potential acquisition” Friedberg ties into his thesis remains uncertain.
• Ellison‑linked involvement in TikTok: In September 2025, an Oracle‑led consortium fronted by Larry Ellison agreed to take roughly an 80% controlling stake in TikTok’s U.S. business, with Oracle running U.S. data and the recommendation algorithm under a new American entity. (reuters.com) Reporting has already framed Ellison as emerging as a major media‑tech power player, connecting his backing of Skydance/Paramount with his role in TikTok US. (forbes.com)

However, Friedberg’s claim is that this combination will, over multiple years, “rewrite the entire media landscape” with new dominant players and business models centered on the Ellison empire. Only a few months have elapsed since both the Paramount Skydance merger and the TikTok deal framework; the Warner Bros. Discovery piece is not yet resolved, and the long‑term competitive and business‑model impacts are inherently not observable over such a short window. Thus, while early moves align with the setup of his thesis, whether they actually lead to the sweeping, Ellison‑centric restructuring he describes cannot yet be determined.

As of November 30, 2025, neither of the specific outcomes Friedberg described has occurred, but the time window (“within the next several years”) has not expired.

What the prediction requires

1. (a) A corporate merger between Ellison‑controlled legacy media assets and TikTok US. In practice this would mean Paramount/Skydance (and possibly a future Warner Bros. Discovery acquisition) combining corporate structures with TikTok’s U.S. entity.
2. (b) Or a deep commercial integration where TikTok itself sells premium studio content (e.g., HBO/Discovery shows) inside the TikTok app on a paid basis (subscription or per‑episode).

What has actually happened so far

• David Ellison completed the merger of Skydance Media with Paramount Global in August 2025, forming Paramount Skydance, which he controls as chairman and CEO.(reuters.com) This matches the “Ellison‑controlled legacy media” setup assumed in the prediction.
• A Trump‑brokered restructuring of TikTok’s U.S. operations would give an Oracle‑led U.S. investor group major control, including licensing TikTok’s recommendation algorithm and overseeing security. Public reporting indicates Larry Ellison/Oracle and allied media tycoons (including the Murdochs) as key stakeholders, with David Ellison frequently mentioned in connection with the prospective board.(theguardian.com) This creates common ownership and influence, but it is not a corporate merger of Paramount Skydance (or WBD) with TikTok US.
• Paramount Skydance is actively pursuing a takeover of Warner Bros. Discovery (the owner of HBO and Discovery), but as of late November 2025 WBD has rejected multiple bids and no deal has closed.(en.wikipedia.org) So the “Ellison‑controlled Warner Bros. Discovery” part of the scenario is still hypothetical.
• TikTok has launched and expanded its Series product, which lets individual creators sell premium, paywalled long‑form videos (up to 20 minutes) directly in the app.(newsroom.tiktok.com) However, there is no public evidence that Warner Bros. Discovery (HBO/Discovery) or Paramount Skydance are using Series or any other TikTok product to sell full HBO or Discovery shows as an in‑app subscription or per‑episode offering. Existing WBD–TikTok activity is promotional/social (e.g., memes, hashtag challenges, marketing clips), not full catalogue distribution.(businessinsider.com)

Why the verdict is ‘inconclusive’

• Outcome (a) has not happened: there is no announced corporate merger between TikTok US and Paramount Skydance or (yet) WBD—only overlapping ownership/influence via the Ellison/Oracle group.
• Outcome (b) has not happened: TikTok does offer paywalled premium content via Series, but not as a distribution channel for HBO/Discovery shows in the way Friedberg described.
• Because the prediction explicitly allows for this to occur “within the next several years” and we are only about two months past the podcast date (September 19, 2025), the forecast cannot yet be definitively judged right or wrong.

So, as of late 2025: the structural pieces Friedberg anticipated (Ellisons deeply involved in both a major legacy media conglomerate and TikTok US) are partially in place, but neither the specific merger nor the TikTok‑hosted HBO/Discovery premium offering he described has materialized yet. Hence the status is inconclusive (too early to tell).

As of November 30, 2025, there is not yet an Ellison-controlled, fully integrated "legacy studios + TikTok‑style" video platform whose actual performance can be measured against YouTube and Netflix, so the prediction can’t reasonably be judged.

What has happened:

• David Ellison’s Skydance completed its $8B merger with Paramount Global on August 7, 2025, creating Paramount Skydance and giving the Ellison family control of a major legacy studio and streaming portfolio (Paramount+, Pluto TV, CBS, cable networks, etc.). (prnewswire.com)
• In parallel, an Oracle‑ and Silver Lake‑led consortium (with MGX and other U.S. investors) is in the process of acquiring a majority stake in TikTok’s U.S. operations, with ByteDance reduced below 20%. The deal structure and 90–120‑day closing windows are laid out, but the transaction is still being finalized and TikTok U.S. has not yet been re‑launched under the new ownership or technically integrated with Paramount Skydance’s services. (reuters.com)
• Analysts and reporters describe Larry and David Ellison as building a new media empire positioned to challenge incumbents like Netflix and Disney, potentially including further acquisitions such as Warner Bros. Discovery, but these are characterized as ambitions and strategic positioning rather than an already‑realized platform at scale. (businessinsider.com)

Against this, current viewing‑share data still show YouTube and Netflix as the dominant global video platforms, with YouTube at about 13.4% and Netflix at 8.8% of U.S. TV viewing in July 2025, while Paramount’s streaming portfolio (Paramount+ plus Pluto TV) sits around 1.9–2.2% and Warner Bros. Discovery/Max around 1.5%. (nielsen.com) There is no evidence yet of a combined Ellison-owned service (e.g., Paramount+ / Pluto tightly fused with TikTok distribution and premium upsells) that is directly rivaling YouTube and Netflix in either total global watch time or creator relationships.

Because Friedberg’s prediction was explicitly about what may happen over time once such a convergence occurs, and the necessary assets (especially TikTok U.S.) are only just being assembled and not yet product-integrated or at scale, it is far too early—barely a couple of months after the podcast—to say whether this future Ellison media‑and‑social platform will or will not become a top‑tier competitor to YouTube and Netflix. The enabling deals are incomplete and the hypothesized product does not yet exist in a testable form. Therefore, the outcome is currently inconclusive (too early to call).

Evidence from coverage of the 2025 Las Vegas Grand Prix shows that the prediction came true. A Nevada Business article on The Venetian Resort’s F1 activations for the 2025 Las Vegas Grand Prix (held November 20–22, 2025) reports that the resort “welcomed the hosts of the All-In Podcast – Chamath Palihapitiya, Jason Calacanis and David Friedberg, for an exclusive Las Vegas Grand Prix weekend experience”, noting that they recorded an episode at The Venetian and “took in the excitement of the race at the Velocity Club, the resort’s premier trackside viewing platform.” (nevadabusiness.com)

Since this explicitly places Friedberg and Jason Calacanis at the Las Vegas Formula One event on the 2025 race weekend in November, the prediction that they would be at the Las Vegas F1 event in November 2025 is correct.

This prediction concerns a future event: the Social Security Old-Age and Survivors Insurance (OASI) Trust Fund actually becoming insolvent (unable to fully pay scheduled benefits) sometime between 2030 and 2033.

As of November 30, 2025, that time window has not started yet, so the trust fund has not become insolvent and cannot yet be evaluated against the 2030–2033 prediction window.

Current Social Security Trustees’ reports project that the OASI Trust Fund will be depleted around 2033, after which incoming payroll taxes would still cover roughly three-quarters of scheduled benefits, but full benefits could not be paid without legislative changes. This projection matches the timing mentioned in the prediction, but it remains only a projection until the 2030–2033 period actually occurs and we see what happens in reality.

Because the prediction is about what will occur between 2030 and 2033, and we are still in 2025, there is not yet enough real-world evidence to say whether the prediction is right or wrong.

Therefore, the correct status is: it is too early to tell whether this prediction will come true.

The prediction states that “a first FDA-approved human drug whose primary mechanism of action is Yamanaka-factor-based cellular rejuvenation … will be approved in the United States roughly 7–12 years from 2025, i.e., between 2032 and 2037.” That approval window (2032–2037) has not started yet as of the current date (November 30, 2025). Therefore, it is too early to know whether such a drug will in fact be approved in that time frame. No amount of current web or scientific literature search can confirm or falsify an approval that is explicitly predicted for future years. Hence, the correct status is inconclusive (too early).

The prediction’s timeframe is “within roughly the next few years (by around 2029).” As of November 30, 2025, we are only about three months past the podcast release (August 22, 2025), so it is too early to judge whether this 2029‑ish claim will ultimately be correct.

What we can say so far:

• There is a visible industry trend toward more modular and specialized approaches, e.g.:
  • Mixture‑of‑experts (MoE) architectures and router‑based systems in both research and production settings.
  • Specialized / small language models (often called SLMs or task‑specific models) such as Microsoft’s Phi series and many on‑device models targeting phones and edge devices, marketed as lower‑cost and more efficient per token than frontier LLMs.
• There is also strong economic and regulatory pressure to reduce energy use and cost per token in large‑scale AI deployments (e.g., data‑center power constraints, GPU scarcity, and cloud cost concerns discussed widely in AI industry reporting in 2024–2025).

However, the prediction contains three separate long‑range components, each aimed at around 2029:

1. A significant transition of AI toward architectures built from networks of smaller, specialized models.
2. Achieving roughly 10x–100x reductions in energy and dollar cost per token relative to 2025 levels.
3. A resulting substantial increase in total token production versus 2025.

Current public information (as of late 2025) is not sufficient to:

• Confirm that SLM‑network architectures will dominate over large monolithic frontier models by ~2029.
• Verify an achieved 10x–100x aggregate reduction in cost/energy per token versus 2025 baselines, at production scale across the industry.
• Measure industry‑wide token production in 2029 relative to 2025.

Because the target year (≈2029) has not yet arrived, and the necessary market‑wide metrics are not yet observable, the prediction cannot be definitively evaluated.

Therefore the correct status as of November 30, 2025 is: inconclusive (too early to tell).

The prediction concerns the late 2020s and explicitly ties its timing to the 2028 U.S. election cycle. As of November 30, 2025, that election cycle has not yet begun in earnest, and there is still substantial time for party dynamics and leading figures to change.

While Alexandria Ocasio-Cortez and other democratic-socialist–aligned figures are nationally prominent and influential within the Democratic Party, it is too early to determine whether a clearly socialist-aligned politician (such as AOC) will, by the 2028 cycle, become the major national standard-bearer or leading candidate of the party. The key condition of the prediction is specifically about their status by 2028, not just rising prominence before then.

Because the forecast’s target period has not yet arrived, the prediction cannot be judged as right or wrong at this time. It is therefore inconclusive (too early).

The prediction explicitly concerns what will happen “over the coming years” to self‑identified socialist or highly left‑wing movements in the U.S. The podcast was released on 22 August 2025, and today is 30 November 2025, only a little over three months later. Multi‑year political momentum and support trends (especially for ideological movements) are typically judged over at least one or more election cycles (e.g., 2026 midterms, 2028 presidential), which have not yet occurred. Even if current polls or news showed short‑term increases or declines, that would not be enough to determine whether these movements will continue gaining momentum versus plateauing or declining over the coming years. Therefore, there is not yet enough elapsed time or data to label the prediction as right or wrong.

As of 30 November 2025, only a few months have passed since the prediction was made on 15 August 2025, while the prediction window is explicitly “the next year or two, particularly in the next election cycle.” That window runs roughly until mid‑2027 and includes the 2026 election cycle, which has not yet fully played out. So it is too early to judge whether the issue will become widely popular in national politics.

There are clear signs that limiting institutional purchases of homes is emerging as a policy idea: for example, New York Governor Kathy Hochul has proposed rules restricting hedge funds and private‑equity firms from buying homes early in their listing period and cutting related tax benefits, explicitly targeting “shadowy private equity giants” for worsening housing shortages. (apnews.com) California’s AB 2584 would bar large institutional investors (owning >1,000 single‑family homes) from buying more single‑family properties for rentals, and has advanced through the legislature. (a24.asmdc.org) Washington’s SB 5496 seeks to limit excessive single‑family home buying by entities owning more than 25 such properties. (lawfilesext.leg.wa.gov) A Virginia bill aims to ban large investment firms from purchasing single‑family homes, (axios.com) and Texas has debated but not yet passed measures to curb institutional homebuying. (houstonchronicle.com) These show the idea is gaining traction in various states.

However, the prediction is about becoming a widely popular political issue and policy plank in the next U.S. election cycle, not merely about the existence of some state‑level proposals or growing policy discussion. With the key election cycle (2026) still ahead and the full 1–2‑year horizon not yet elapsed, we cannot yet determine whether it will become a broadly central, bipartisan or mass‑salience campaign issue. Therefore the correct status today is “inconclusive (too early).”

The prediction explicitly refers to "the coming years" and says socialism will "sweep over this nation" (the U.S.), which implies a multi‑year horizon beyond the few months that have passed between the podcast date (August 9, 2025) and now (November 30, 2025). That time frame has clearly not elapsed, so the claim can’t be fairly judged yet.

Substantively, current indicators do not yet show socialism as nationally dominant:

• National polling still shows most U.S. voters prefer a free‑market system over socialism; one 2025 survey finds 71% of likely voters say a free‑market economy is better than socialism and only 12% say socialism is better. (rasmussenreports.com)
• Gallup‑based reporting indicates that while support for capitalism is at a historic low (about 54% viewing it favorably), socialism remains less popular overall, with roughly 39% favorability nationwide. (foxnews.com)
• Within the Democratic Party, socialism has become relatively popular (about two‑thirds of Democrats view it positively), but Republicans and independents remain largely negative, so it is far from a cross‑partisan national consensus. (yournews.com)
• Politically, the U.S. House of Representatives recently passed a resolution explicitly denouncing socialism, with a large bipartisan majority voting in favor—hardly evidence of socialism having already “swept” national politics. (time.com)

However, because the prediction’s horizon is “coming years,” it is too early, as of late 2025, to decide definitively whether socialism will eventually “sweep over” the U.S. and produce the broad unhappiness Friedberg fears. The appropriate classification is therefore inconclusive (too early to tell), even though available data so far do not show the predicted dominance of socialism nationwide.

The prediction specified a window of ~18 months from August 2025, i.e. until around February 2027, for “a Gen‑4 reactor [in the U.S.] deployed or started production.” As of November 30, 2025, that deadline has not yet arrived, so we cannot know whether it will ultimately be met.

Current evidence shows several U.S. Generation IV / advanced reactor projects are under development or early construction, but not yet deployed or producing power:

• Project Pele microreactor (DoD/BWXT) – DoD and DOE report that ground was broken at Idaho National Laboratory in 2024 and that the transportable high‑temperature gas reactor is being built by BWXT, with assembly starting in 2025 and potential operation at INL as early as 2026; it is expected to be the first Generation IV reactor to generate electricity in the U.S. (defense.gov)
• Kairos Power Hermes 1 & 2 (fluoride‑salt‑cooled reactors) – The NRC has issued construction permits for Hermes 1 (non‑power demonstration) and, more recently, for Hermes 2 units that Kairos describes as the first electricity‑producing Gen IV reactors permitted for construction in the U.S., but these units are not yet operating. (powermag.com)
• TerraPower Natrium (sodium‑cooled fast reactor, Wyoming) – Non‑nuclear site work started in 2024, with nuclear construction expected after an NRC construction permit targeted for 2026 and operation around 2030; as of late 2025 it has not begun nuclear island construction or operation. (powermag.com)

These timelines suggest it is plausible that at least one Gen IV project (e.g., Project Pele or a Hermes/Natrium unit) could be under construction or even operating by early 2027, but that outcome lies in the future relative to the current date. Because the 18‑month window has not elapsed and we lack information from beyond November 2025, we cannot yet determine whether Friedberg’s prediction will ultimately be correct.

Therefore, the appropriate classification as of now is “inconclusive (too early).”

By the time of (and even before) the August 1, 2025 episode, there were already multiple AI systems explicitly tuned to particular religious beliefs and available to the public:

• Hindu-specific: GitaGPT is a GPT‑3–based chatbot trained on the Bhagavad Gita to answer life questions using its verses, described as a ChatGPT‑like platform for “Gita ka gyaan” (wisdom from the Gita).(timesofindia.indiatimes.com)
• Islam-specific: A range of bots such as QuranGPT, HadithGPT, and Islam & AI focus on answering questions from the Quran and/or Hadith; a 2023 feature explicitly refers to such systems as “religious AI chatbots,” noting they are trained on Islamic scriptures and used by tens of thousands of users across many countries.(thenationalnews.com)
• Christian-specific: Multiple BibleGPT products exist, including a web service positioning itself as an “AI Bible companion” that provides scripture‑rooted answers, and an iOS/Android app BibleGPT: Study Bible with AI that markets “AI-powered answers to your Biblical and Christian questions,” offered with free and paid tiers on mainstream app stores.(biblegpt-la.com)
• Catholic-specific: Magisterium AI, created by Longbeard, is explicitly branded as “Catholic AI,” built on a curated database of about 27,000 Catholic documents and marketed as a tool to answer questions on Catholic doctrine; it has tens of thousands of monthly users and is integrated into the commercial Hallow prayer app.(washingtonpost.com)

These systems are all commercially or publicly available AI tools, intentionally trained or constrained to specific religious corpora and faith traditions, matching the idea of “religious AI” tuned to users’ beliefs. Because such products clearly existed (and have grown more numerous) by late 2025, the prediction that we would “see AI that’s tuned to people’s religious beliefs” has, in substance, come true.

As of November 30, 2025, there is no evidence that any jurisdiction has formally granted a professional license (such as bar admission or a medical license) directly to an AI system or model.

Medicine:

• Some AI systems have passed or achieved high scores on medical licensing exams (e.g., the Xiaoyi robot in China passing the national medical licensing exam; OpenEvidence’s model scoring up to 100% on the USMLE), but these systems are not recognized as licensed physicians. They function as decision-support tools for human doctors, not as independently licensed professionals. (en.wikipedia.org)
• U.S. regulators are explicitly reinforcing that only humans can be licensed clinicians. California’s AB 489, signed into law in October 2025, prohibits AI systems from presenting themselves as licensed health professionals and extends existing title-protection rules to AI developers and deployers. The California Medical Board has emphasized that only a “natural person” may be a licensed physician in the state. (medscape.com)

Law:

• In the U.S., unauthorized practice of law (UPL) statutes in all states restrict legal practice to individuals who are licensed attorneys, and existing case law and guidance frame the practice of law as limited to natural persons who have been admitted to the bar. (americanbar.org)
• While LLMs have passed bar-exam-style benchmarks, and legal AI tools like Harvey or others are widely used, they operate under the supervision of human lawyers rather than holding any bar license themselves. Regulatory reform efforts in states like Utah, Arizona, and Washington expand business models and tech use but still premise actual legal practice on human licensees. (reuters.com)

Emerging proposals, not law:

• The U.S. Healthy Technology Act of 2025 (H.R. 238) would allow AI/ML systems to qualify as practitioners eligible to prescribe drugs under specified conditions, but as of late November 2025 it remains only introduced and referred to committee, with no passage in either chamber. (congress.gov)

Given this:

• The specific event Friedberg predicts — AI models being formally allowed to obtain professional certifications like bar admission or medical licensing, to a degree that a Polymarket contract could cleanly resolve on that fact — has not occurred by November 30, 2025.
• However, his prediction is explicitly open-ended (“there will be a date when this happens”) with no time horizon, so the fact that it has not yet occurred does not make it wrong; it is simply untested so far.

Because the outcome is about an unspecified future date and there is still ample time for such regulatory changes to (or not to) occur, the correct classification today is “inconclusive (too early)”, not right or wrong.

The prediction explicitly concerns what will happen “over the next several years” if U.S. AI regulation remains a fragmented, state‑level patchwork. As of November 30, 2025, that time horizon has not elapsed, and both the regulatory structure and its long‑term economic effects are still evolving.

Current situation (late 2025):

• The U.S. does not yet have a comprehensive federal AI statute; policy is a mix of federal executive actions (e.g., the 2023 AI Executive Order and subsequent guidance) plus growing state‑level laws and bills (such as state AI, data‑privacy, and automated decision‑making regulations).
• Several states have passed or are considering AI‑related rules (e.g., focused on deepfakes, hiring algorithms, or sector‑specific uses), and commentators already describe this as an emerging “patchwork” of state AI and algorithmic regulations.
• However, there is no clear, widely documented evidence yet that this patchwork has made it “practically impossible” for major providers like Google or OpenAI to serve customers across state lines, nor that it has materially harmed U.S. consumers or measurably slowed AI‑related job growth nationwide. Major AI services (e.g., large cloud AI platforms and popular consumer AI tools) remain broadly available across states, and AI‑related hiring and investment remain strong by late 2025.

Because:

1. The specified timeframe (“over the next several years”) has not passed, and
2. The conditional (“if regulation remains fragmented at the state level”) and outcome (nationwide, material impediment and job‑growth harm) concern medium‑ to long‑term structural effects that cannot yet be robustly measured,

it is too early to determine whether the prediction ultimately proves correct or incorrect.

So, based on evidence available by November 30, 2025, the accuracy of this prediction is inconclusive (too early to tell).

News coverage and retrospective summaries show that the Musk–Trump relationship followed the arc Friedberg described.

After Musk savaged Trump’s “One Big Beautiful Bill” as a “disgusting abomination” and even floated creating an “America Party,” Trump threatened Musk’s contracts and mused about deporting him, producing a very public rupture between MAGA and its most visible tech ally.(washingtonpost.com) However, Musk later deleted his most incendiary posts (including Epstein insinuations) and publicly expressed regret, while Trump began calling Musk a “good person” who’d had a “bad moment,” signaling a deliberate de‑escalation.(en.wikipedia.org)

By August 2025, Musk was again praising Trump and Vice President JD Vance and explicitly backing Trump’s law‑and‑order moves in Washington, which contemporaneous coverage framed as a “possible reconciliation” and renewed alignment with the MAGA project.(thedailybeast.com) The Trump–Musk feud entry now characterizes the overall status as “Reunited” and notes that relations “began to thaw” as Musk reshared Trump posts and Trump publicly vouched for him.(en.wikipedia.org)

The clearest confirmation of Friedberg’s “mutual dependence” language came at Charlie Kirk’s September 21 memorial: Trump and Musk sat together in their first joint appearance since the blow‑up, and multiple reports recount Musk telling Trump, “If we don’t hang together we will surely hang separately,” and saying he was “proud to sit with my president tonight” while calling for GOP unity.(nypost.com) Business Insider likewise reports Trump insisting Musk “doesn’t have a choice” but to side with the GOP and that MAGA leaders view his tech influence as vital for their future success—an explicit acknowledgment of mutual reliance between the political and tech camps.(businessinsider.com)

There are still substantive policy tensions, but by late 2025 the feud has clearly cooled and the two sides have publicly re‑aligned around shared political goals, matching Friedberg’s prediction.

As of November 30, 2025, only about 5 months have passed since the July 4, 2025 episode, while the prediction’s horizon is ~18 months (into early 2027), so the full time window has not elapsed.

What we do know so far:

• Harvard is under material financial pressure. Federal funding freezes and cuts totaling roughly $2.6B have contributed to a $113M operating deficit in FY 2025, prompting borrowing, hiring freezes, and cost cuts, but not an existential inability to operate. (reuters.com)
• Harvard Management Company is planning/has agreed to sell about $1B of private equity fund stakes (roughly 5% of its PE portfolio) via a secondary sale advised by Jefferies, with Lexington Partners as buyer. The process began in 2024 and is described as part of broader liquidity management rather than a sudden, last‑ditch move driven solely by the 2025 funding crisis. (thecrimson.com)
• Reporting on the broader secondary market indicates that university PE stakes (including Harvard and Yale) are generally selling at moderate discounts, with typical LP secondary sales around an 11% discount to NAV and Yale’s large deal reportedly at a <10% discount—well below the 20–40% haircut specified in the prediction. Harvard-specific pricing has not been publicly disclosed. (fortune.com)

Key parts of Friedberg’s forecast therefore cannot yet be confirmed or falsified:

• It is too early to say whether by early 2027 Harvard will face a severe budget shortfall so acute that it “will not have the budget to sustain themselves.”
• While Harvard is indeed selling private equity stakes, current reporting frames this as a strategic liquidity move rather than a forced, fire‑sale situation with the university’s “back totally against the wall.” (thecrimson.com)
• There is no public evidence so far that Harvard’s PE secondary sales have cleared at the 20–40% discounts to reported NAV that the prediction specified; available benchmarks suggest materially smaller discounts.

Because (1) the 18‑month window has not expired and (2) the critical discount and severity conditions of the prediction are not yet observable or disclosed, the correct classification at this time is **“inconclusive (too early)” rather than right or wrong.

As of November 30, 2025, none of the revenue years in the prediction have fully occurred:

• OpenAI 2025 revenue (~$13B): Calendar year 2025 has not ended, and OpenAI is a private company. Only leaked or projected run-rates for 2025 exist so far (e.g., various reports of multi‑billion 2024–2025 run‑rate), but no final, audited 2025 revenue is available yet, so we cannot say whether “approximately $13B” is right or wrong.
• OpenAI 2029 revenue (~$125B): Calendar year 2029 is entirely in the future.
• Anthropic 2027 revenue (~$35B): Calendar year 2027 is also in the future.

Because the relevant years (2025, 2027, 2029) are not yet completed (and in two cases haven’t even started), the prediction cannot currently be evaluated. Therefore the correct classification is "inconclusive (too early)".

Friedberg’s prediction is explicitly about a multi‑year (roughly 1–3 year) outcome after Figma’s IPO, not just the first few months of trading. Since Figma IPO’d on July 31, 2025 on the NYSE under ticker FIG at $33 per share, only about four months have elapsed as of late November 2025—well short of the 1–3 year window he described.(economictimes.indiatimes.com)

We can say that so far the relative-value call has been directionally correct: Figma is trading around the mid‑$30s (about $36 per share recently), while Adobe has fallen from roughly $358 at the July 31, 2025 close to around $320 now.(stockanalysis.com) A market‑neutral trade put on at IPO pricing (long FIG at ~$33, short ADBE at ~$358, dollar‑for‑dollar) would currently show a gain on both legs, implying a profitable spread so far. But the claim was that this would be a good trade over the next several years, i.e., over the full 1–3 year post‑IPO period, and that longer horizon has not yet played out.

Because there is still substantial time left in the stated 1–3 year window and the trade’s ultimate performance over that full period cannot yet be known, the prediction’s truth value is inconclusive (too early to tell).

Open sources show that after the June 2025 U.S. and Israeli strikes on Iran’s nuclear and military infrastructure, Israel did not pivot to a quiet, status‐quo de‑escalation with the Iranian regime, but instead maintained and even emphasized activities and messaging aimed at weakening it.

1. Stated goal of weakening or toppling the regime. Reporting during and after the war makes clear that Israeli leadership framed the June campaign not just as non‑proliferation, but as an effort to strike at the foundations of Ayatollah Khamenei’s rule and even pursue regime change. Reuters described Israel’s air campaign as aiming to “severely weaken or even collapse” Khamenei’s government, going beyond narrow nuclear objectives.【7news13】 A separate analysis reported Netanyahu explicitly urging Iranians to “rise” against an “evil and oppressive regime” and presenting the strikes in the context of supporting a different future for Iran, i.e., overt regime‑destabilization rhetoric rather than narrow deterrence.【7news14】 Post‑war diplomacy coverage likewise notes Netanyahu pushing for dismantling Iran’s regime and resisting U.S. moves toward renewed talks, indicating an ongoing strategic objective rather than a one‑off strike campaign.【6news14】

2. Information and psychological operations directed at Iranians. During and after the June war, Israel markedly escalated Persian‑language outreach and influence operations targeting the Iranian public. An analysis of Israel’s “digital campaign” documents how official IDF and Foreign Ministry Persian‑language accounts used monarchist symbols (the lion and sun) and messaging tailored to anti‑regime audiences, explicitly appealing to segments of the Iranian population opposed to the Islamic Republic.【7search0】 Another report describes Israel hacking Iran’s state broadcaster IRIB during the conflict to air a message urging Iranians to rise up against the regime, a direct attempt to incite internal opposition rather than simply degrade military capabilities.【7search4】 The IDF also used its Persian account to highlight Mossad’s secure contact portal and effectively invited Iranians to provide information or collaborate, an unusually overt attempt to recruit assets inside Iran.【7search2】 These are precisely the kinds of ongoing Mossad/IDF efforts to destabilize the regime that Friedberg was speculating about.

3. Continued posture and new influence initiatives after the ceasefire. The formal ceasefire of June 24, 2025 ended large‑scale open hostilities, and has broadly held in terms of direct kinetic strikes between Israel and Iran.【5search18】 But Israeli officials openly signaled that the conflict with the regime was not over. In September, Mossad chief David Barnea publicly boasted that Mossad retains “operational capabilities” even “in the heart of Tehran” and emphasized that the agency would continue to act against Iran, underscoring an ongoing covert campaign rather than a pause or reset.【2news14】 In November 2025, the Jerusalem Center for Security and Foreign Affairs launched a new Persian‑language influence campaign explicitly aimed at exposing “disinformation mechanisms of the Ayatollah regime” and fostering dialogue about freedom, human rights, and corruption among Iranians—an Israeli civil‑society effort clearly aligned with eroding the regime’s legitimacy rather than accepting a stable status quo.【7search3】

4. Outcome in Iran: crisis, not a peaceful transition. Domestically, Iran has moved into what analysts describe as a 2025 “internal crisis,” characterized by mass arrests, accelerated executions, and heightened regime paranoia after the war.【6search17】 That trajectory—escalating repression, continued Israeli anti‑regime rhetoric and influence operations, and no normalization between Jerusalem and Tehran—matches Friedberg’s core claim that this would not be a “quiet, peaceful transition” for the Islamic Republic after the June strikes, but an environment in which Israel kept trying to weaken whatever regime remained.

We do not have open-source confirmation of new, post‑ceasefire kinetic Mossad/IDF attacks inside Iran itself; such covert actions may or may not have occurred and would often be deniable. However, the observable record—Israeli leadership’s articulated regime‑change aims, continued psychological and influence operations in Persian, and public assurances by Mossad’s chief that operations inside Iran will continue—shows that Israel did indeed persist in efforts directed at undermining the current Iranian regime, rather than settling into a quiet de‑escalation. On that basis, Friedberg’s prediction is best judged as directionally right, even if the exact form of the “next shots” has so far been more informational and political than additional large‑scale bombing inside Iran.

Friedberg’s prediction had two key components:

1. Starting point: that the June 2025 New York City mayoral Democratic primary would mark the beginning of a wave of democratic socialist / left‑populist victories in major U.S. cities.
2. Trajectory: that over the next several election cycles there would be a growing wave of such candidates winning major municipal offices.

What has happened so far (through Nov 30, 2025):

• On June 24, 2025, Assemblymember Zohran Mamdani, a self‑identified democratic socialist and member of the Democratic Socialists of America (DSA), won the New York City Democratic mayoral primary. 【0search18】 He is explicitly described as a democratic socialist and DSA member. 【1search18】【1search0】
• On November 4, 2025, Mamdani won the general election and is now mayor‑elect of New York City, the largest U.S. city, running on a democratic socialist platform. 【3search19】
• In the 2025 Seattle mayoral election, progressive activist Katie Wilson defeated incumbent Bruce Harrell; she is widely described as a democratic socialist or self‑proclaimed socialist and ran on a left‑populist platform focused on affordability, homelessness, and taxing the wealthy. 【3search24】【2news18】【2news13】【2search0】
• A running list of elected socialist mayors in the U.S. now includes NYC and Seattle (terms beginning 2026) alongside earlier DSA‑aligned mayors in smaller or mid‑sized cities, indicating that socialist mayors exist but only some of these wins occurred after June 2025. 【1search20】

Why the prediction is not yet judgeable:

• The “beginning” part of his claim is broadly consistent with events: a DSA democratic socialist winning the NYC primary (and then the mayoralty) in June–November 2025, followed by a democratic socialist winning in another major city (Seattle) the same cycle, does look like the start of a left‑populist surge in big‑city municipal politics.
• However, he explicitly forecast a “growing wave over the next several election cycles.” As of November 30, 2025, we have seen only one main municipal election cycle (the 2025 local elections). We don’t yet have evidence from multiple future cycles to confirm whether this is a sustained, growing trend or a one‑cycle peak.

Because the time horizon (“next several election cycles”) has not passed and we cannot yet see whether the trend grows, stalls, or reverses, the prediction cannot be definitively rated as right or wrong at this point. It is directionally supported so far but still too early to call, so the appropriate status is inconclusive.

The prediction is explicitly time‑bounded to occur “within the next few U.S. election cycles.” As of the current date (November 30, 2025), we are only partway between the 2024 and 2026 U.S. federal election cycles. That is at most one cycle into what is usually interpreted as “a few” (commonly 2–3 cycles), so the stated timeframe has not elapsed.

Additionally, the claim is about how figures like Kamala Harris will be perceived relative to an “increasingly socialist” youth/urban wing of the Democratic Party. This is a long‑horizon, qualitative shift in intra‑party ideology and public perception, not something that can be definitively judged just months after the prediction. Even if some commentary today frames establishment Democrats as more centrist than the party’s left flank, that dynamic pre‑dated the prediction and does not yet show a clearly new, widely accepted reclassification of Harris herself as a “conservative candidate” within Democratic politics.

Because (1) the full timeframe (“next few election cycles”) has not passed, and (2) any definitive, widely accepted perception change of the described magnitude has not been clearly documented or testable yet, this prediction cannot be fairly evaluated at this time.

David Friedberg made this comment about Zohran Mamdani’s upset win in the NYC Democratic mayoral primary on the June 28, 2025 episode of the All-In podcast, calling it “the beginning of a wave that will sweep over [America].”(speakai.co) The user’s normalized prediction frames this as a decade‑long shift toward majority support for socialist‑style policies across much of the U.S. As of today (November 30, 2025), barely five months have passed since the prediction, and the relevant time horizon—“over the next decade,” i.e., through roughly 2035—has not elapsed. Because this claim is about long‑term national political realignment rather than immediate election outcomes, there is not yet enough elapsed time or data to fairly determine whether such a broad wave has materialized or failed, so the prediction must be considered too early to judge.

As of 30 November 2025, the FOXO3-based age‑reversal work that Friedberg was referring to is still at the primate/preclinical stage. The key 2025 Cell paper by Lei et al. reports systemic age‑reversal effects in aged macaques using FOXO3‑enhanced, senescence‑resistant mesenchymal progenitor cells (SRCs) and attributes much of the effect to exosome‑mediated mechanisms, but it is entirely in non‑human primates and in vitro models, not humans. (sciencedirect.com) Follow‑on commentaries in Cell Regeneration and related outlets explicitly frame this work as a proof‑of‑concept that positions FOXO3‑enhanced MPCs/SRC‑exosomes as promising candidates for future clinical anti‑aging interventions, again without mentioning any ongoing human trials. (cellregeneration.springeropen.com) Popular and expert summaries likewise describe the therapy as early‑stage and “years away from real‑world use,” underscoring that translation to human trials has not yet happened. (biohackingnews.org) A targeted search of trial registries and the scientific literature through late 2025 finds FOXO3 in basic/translational contexts but no registered human clinical trials testing FOXO3‑engineered stem cells or their exosomes as systemic age‑reversal therapies analogous to the macaque study. Instead, all detailed reports match the same preclinical FOXO3‑SRC primate work. (pubmed.ncbi.nlm.nih.gov)

However, Friedberg’s prediction is about the situation “over the next couple of years”—i.e., roughly by 2027. Since 2027 has not yet arrived, we cannot know whether such human trials will in fact be underway by that time, even though they are not underway yet in 2025. Therefore the prediction cannot be definitively graded now and is best classified as inconclusive (too early) rather than right or wrong.

As of November 30, 2025, no injectable or infusion-based age-reversal therapy based on stem cells or Yamanaka-style partial reprogramming is approved or commercially available to average consumers; such approaches remain preclinical or at best in very early, disease-specific human trials (e.g., Life Biosciences’ planned 2025 optic-nerve trial and YouthBio’s preparatory steps toward an Alzheimer’s trial). (p05.org) Major reviews note that partial reprogramming with Yamanaka factors has extended lifespan in mice but has not yet entered human trials for aging itself, and aging is not recognized by regulators such as the FDA as a disease indication that can be directly targeted, further slowing development of any broad age-reversal therapy. (longevity.technology) Medical/science overviews likewise emphasize that, as of mid‑2025, there are no approved Yamanaka- or MPTR-based anti-aging treatments available to the public; any offerings are experimental or speculative. (mededgemea.com) However, Friedberg’s prediction is about whether such a therapy will exist by roughly 2035, so with nearly a decade remaining on that timeline, there is not yet enough information to say whether it will ultimately be right or wrong.

The prediction’s own timeframe is “in the near term,” which we normalized as within a few years from June 2025. As of November 30, 2025, only about five months have passed, so the window for judging it has clearly not expired.

Substantively, there are early signs consistent with the prediction:

• Huawei is mass‑shipping Ascend 910C AI accelerators built on SMIC’s domestic 7 nm process as an alternative to Nvidia’s H100/H20 in China; these chips reach roughly 60% of an H100’s performance and are being deployed in large CloudMatrix systems.(mihutz.com)
• Huawei has announced the Ascend 920 on SMIC’s 6 nm node, aiming to exceed Nvidia’s China‑specific H20 in raw performance, though independent benchmarks and large‑scale deployment are still pending.(tomshardware.com)
• U.S. export controls and recent Chinese regulatory moves limiting new Nvidia deployments inside China are pushing domestic firms toward local AI silicon (e.g., Huawei Ascend), creating the beginnings of a structural, China‑based challenger in that market.(reuters.com)

However, as of late 2025 these developments have not yet translated into a clear, global “significant competitive threat” to Nvidia’s overall AI‑chip dominance:

• Analyses consistently find Huawei’s Ascend line still one generation behind Nvidia in performance per watt and ecosystem maturity; training on Ascend often costs more and carries higher integration risk versus Nvidia’s CUDA‑based stack.(ainvest.com)
• SMIC’s 7 nm and 6 nm processes lag TSMC’s leading‑edge nodes that fabricate Nvidia’s top chips, and Huawei’s 2025 production volumes (a few hundred thousand advanced AI chips) are far below Nvidia’s global scale, limiting immediate competitive impact outside China.(evervolve.com)

Overall, China has started developing advanced AI chips on domestic manufacturing processes that challenge Nvidia within China, but it is too early in the stated multi‑year window to say whether this will mature into the “significant competitive threat” to Nvidia’s global dominance that the prediction describes. Hence the status is inconclusive (too early) rather than clearly right or wrong.

The prediction explicitly sets a long horizon: performance dispersion in the S&P 493 “over roughly the next decade.” As of now (late November 2025), only a few months have passed since the June 21, 2025 podcast release date, far too little time to evaluate whether:

• AI adoption has persistently driven “very large performance dispersion” across the S&P 493, and
• this has in fact produced “some of the largest relative‑return opportunities for public‑equity investors in decades.”

While there are already anecdotal signs of differing AI strategies and some early market enthusiasm around AI-forward companies, that short-term behavior cannot conclusively confirm or refute a decade-scale claim about structural return dispersion and opportunities. We would need many more years of realized returns and clear attribution to AI-driven operational changes to make a fair assessment.

Because the forecast window is still in its very early phase, the correct classification at this time is “inconclusive (too early)”.

The prediction explicitly uses a 5–10 year horizon for these newly public companies to compound revenues around 25% annually and become the next generation of "compounders." As of November 30, 2025—about five months after the June 21, 2025 podcast—CoreWeave, Circle, and Chime have only a short public track record, even though early growth looks strong (e.g., CoreWeave’s 2024 revenue grew about 737% year over year and it guides to multi‑billion‑dollar 2025 revenue; Circle reported 53% year‑over‑year revenue growth in Q2 2025, its first quarter as a public company; Chime’s revenue grew roughly 30% in 2024 and continues to rise post‑IPO). (fool.com) Whether many of these firms will sustain ~25% annual growth over an entire 5–10 year period and truly become durable compounders cannot yet be determined, so the prediction is too early to judge.

The prediction explicitly covers “the next five years” from June 21, 2025, i.e., roughly through mid‑2030. As of today (November 30, 2025), less than a year of that five‑year window has elapsed.

Even if we could already observe some dispersion between AI winners and laggards in the Mag 7 or the broader S&P 500, that would not be enough to validate the claim that:

1. This entire five‑year period is unusually good for active stock pickers, and
2. The key driver of that persistent dispersion is AI adoption.

Both points require multi‑year realized performance data (active manager excess returns, cross‑sectional dispersion, sector/industry attribution to AI themes, etc.) that simply do not exist yet for the full horizon. Because the evaluation date (mid‑2030) has not arrived, the prediction cannot be judged right or wrong at this time.

As of November 30, 2025, only about five months of the 1–2 year window (June 21, 2025 through at least June 21, 2026, and possibly June 21, 2027) have elapsed, so the full period the prediction covers has not yet played out.

Evidence so far is supportive of Friedberg’s view:

• High‑quality tech IPOs have occurred and been very well received. Figma went public on July 31, 2025 at $33 per share and raised roughly $1.2 billion; the stock closed its first day over $115, more than 250% above the IPO price, signaling extremely strong demand for a marquee growth software name. (forbes.com)
• Other “fantastic” growth tech/fintech/AI names have also listed with big pops. Circle Internet Group’s June 2025 IPO and CoreWeave’s March 2025 IPO are cited alongside Figma as standout tech offerings; Circle’s first‑day gain was about 165%, while CoreWeave and other AI‑infrastructure/tech IPOs have delivered triple‑digit returns post‑listing, indicating robust investor appetite for top‑tier growth tech issuers. (finance.yahoo.com)
• Broader US IPO data suggests the market is ‘open for business’ in 2025, especially for quality names. By mid‑2025, U.S. IPOs were averaging first‑day pops of around 27–36%, near decade highs, with tech leading the way and 241 U.S. IPOs year‑to‑date—evidence of an active, receptive market. A February 2025 Renaissance Capital review also noted that most sizable IPOs were trading above offer, even if a few large deals dragged down weighted averages. (finance.yahoo.com)

However, Friedberg’s prediction is explicitly about the market remaining open over the next 1–2 years. Conditions could still change (e.g., macro downturn, rate shock, regulatory or geopolitical events) between now and mid‑2026/2027 that might shut the window for high‑quality growth tech IPOs. Because that future part of the timeframe has not yet occurred, we cannot definitively say the prediction has proven right or wrong.

Therefore the correct status at this time is inconclusive (too early): early developments are consistent with his thesis, but the full 1–2 year horizon has not yet passed.

The prediction is explicitly about “the coming years” of the AI transition, implying a multi‑year horizon for evaluating whether a long/short strategy (short S&P 500, long a few emerging AI “category killers”) outperforms simple passive S&P 500 exposure.

The episode containing this remark was released on June 21, 2025.(tapesearch.com) As of today, November 30, 2025, less than one year has passed, which is well short of a reasonable interpretation of “years” (at least ~2–3 years). In addition, the prediction does not specify:

• which exact AI “category killer” stocks to hold,
• the position sizing and rebalancing rules,
• or the precise evaluation window.

Because (1) the stated timeframe has not elapsed, and (2) the strategy is too loosely defined to test rigorously so far, there is not enough information or elapsed time to determine whether Friedberg’s claim about this strategy being unusually attractive and likely to outperform the S&P 500 has proven correct or incorrect.

So the status as of November 30, 2025 is “inconclusive (too early)”.

The prediction window is about five years from the podcast date of June 21, 2025, so it runs roughly until mid‑2030. As of November 30, 2025, we are only a few months into that window, so it is too early to judge whether Microsoft’s eventual total headcount will end up below its recent peak (~250,000 employees).

Recent public data suggests Microsoft’s headcount has been near or above that quarter‑million level in the mid‑2020s, but the prediction concerns where it will settle by 2030, not where it is in 2025. Because that future outcome cannot yet be observed, the correctness of the prediction cannot be determined at this time.

Therefore, the status of the prediction is inconclusive (too early).

The prediction specifies a five‑year horizon starting from around the podcast release date of June 21, 2025. That window runs approximately until June 21, 2030. As of today (November 30, 2025), only a little over five months of that period have elapsed, far short of the full five‑year horizon required to judge whether a portfolio concentrated in AWS, Azure (Microsoft), and GCP (Google) has outperformed the rest of the market. Because the prediction is explicitly long‑dated and we are still early in the evaluation period, it cannot yet be determined whether it is right or wrong.

It’s too early to determine whether this prediction is right or wrong.

The claim is about what will happen by 2032: that the U.S. Social Security system will become functionally insolvent (e.g., trust fund depletion) and that this will in turn trigger higher taxes and/or inflationary policies shifting the burden to younger generations.

As of the current date (November 30, 2025), 2032 is still seven years in the future, and the key event in the prediction (reaching functional insolvency by that date and the specific policy response) has not yet occurred or failed to occur. While current projections by the Social Security Trustees and independent analysts often estimate OASI trust fund depletion around the early–mid 2030s, those are forecasts, not outcomes; we cannot yet say whether the system will in fact be functionally bankrupt in 2032 or whether the specific policy response (over-taxation and/or inflationary burden on younger generations) will happen in the way described.

Because the prediction is explicitly time-bound to 2032 and we are still in 2025, the correct classification is “inconclusive” (too early).

The “Big, Beautiful Bill” they were discussing is the House GOP reconciliation package that became the One Big Beautiful Bill Act (H.R. 1, 119th Congress), signed into law on July 4, 2025.

Observed deficits so far: For the first full fiscal year in this Trump term (FY 2025, ending Sept. 30, 2025), the federal budget deficit was about $1.8 trillion, essentially flat versus FY 2024 and well below $2.5 trillion. (reuters.com) That is after the bill’s passage, so the realized deficit has not yet “climbed to $2.5 trillion” annually.

Forward-looking projections under current law and the bill: Before the bill, CBO’s January 2025 baseline already projected deficits rising from $1.9T in 2025 to about $2.6T by 2034–2035. (americanactionforum.org) After enactment, CBO scoring (summarized by the American Action Forum) found the One Big Beautiful Bill increases cumulative deficits by about $4.1T over 2025–2034, with annual deficits rising to about $3.0T in 2034, roughly $400B higher than the pre‑bill baseline in that year. (americanactionforum.org) A separate analysis from the Committee for a Responsible Federal Budget puts post‑bill deficits over 2026–2035 about $1T higher than CBO’s earlier baseline, with 2035 around $2.6T. (reuters.com)

Implication for the prediction: Friedberg’s normalized claim is that, under this bill and current trajectory, the deficit will rise to roughly $2.5T per year and then stay around that level for multiple consecutive years. As of November 30, 2025:

• We have only one completed fiscal year after the bill’s introduction and only a few months after its enactment.
• The realized deficit is ~$1.8T, not ~$2.5T.
• Long‑term projections do show deficits eventually approaching or exceeding $2.5T in the 2030s, but those years have not occurred yet, and projections could change.

Because the core of the prediction concerns the medium‑/long‑run level of the deficit in years we have not yet reached, there isn’t enough realized data to say whether the U.S. will, in fact, sustain ~$2.5T annual deficits for “the foreseeable future.”

Therefore, the status of this prediction is inconclusive (too early to tell).

1. Condition on the bill being enacted:

   • The House tax-and-spending package at issue is the "One Big Beautiful Bill Act" (H.R. 1), whose tax provisions were developed jointly by House Ways and Means Chair Jason Smith and Senate Finance Chair Mike Crapo; reporting at the time explicitly describes the tax section as a Smith–Crapo effort. (youngkim.house.gov)
   • That bill, with only negotiated modifications around the edges, passed the House on May 22, 2025, passed the Senate on July 1, 2025, and was signed into law by President Trump on July 4, 2025. (en.wikipedia.org)
     → It is reasonable to treat Friedberg’s condition (the Smith/Crapo House tax bill being enacted “substantially as proposed”) as having been met.

2. Did the bill materially reduce the deficit?

   • Multiple analyses of the enacted One Big Beautiful Bill find that it increases, rather than reduces, deficits relative to CBO’s January 2025 baseline:
     • CBO’s static score of the House-passed version estimated about $2.4 trillion higher cumulative deficits over 2025–2034. (nypost.com)
     • A later CBO letter analyzing the enacted law finds it will raise deficits by $4.1 trillion over 2025–2034 (including interest), taking the annual deficit from about $1.9T in 2025 to roughly $3.0T by 2034. (americanactionforum.org)
     • The Committee for a Responsible Federal Budget similarly estimates post‑bill deficits over 2026–2035 will be about $22.7T, roughly $1T worse than prior projections. (reuters.com)
   • So the specific sub‑claim “the bill ultimately yields no real change in the annual deficit / will not materially reduce the deficit” is clearly supported by the best available evidence and is, at this point, effectively correct in direction.

3. Did the deficit rise to about $2.5T per year on a sustained basis?

   • Friedberg’s stronger claim is that the annual deficit would climb to roughly $2.5 trillion per year on a sustained basis in subsequent years.
   • As of late 2025, we do not yet observe such realized deficits:
     • CBO’s January 2025 baseline (before the bill) projected the deficit at $1.9T in 2025, rising gradually toward about $2.6–2.7T by 2034–2035. (edwardconard.com)
     • Post‑enactment analysis of the One Big Beautiful Bill suggests higher future deficits—around $3.0T in 2034, relative to the baseline. (americanactionforum.org)
   • Some commentary summarizing CBO’s outlook notes projected deficits of about $2.4T in 2033, $2.5T in 2034, and $2.7T in 2035, i.e., in the ballpark of his “$2.5T per year” description. (foxnews.com)
   • However, these are forecasts for the 2030s, not outcomes. As of November 30, 2025, only the 2025 fiscal year is underway, with the deficit still around the ~$1.8–1.9T level; the multi‑trillion‑dollar deficits Friedberg foresaw have not yet actually occurred.

4. Why the verdict is ‘inconclusive’:

   • We have strong evidence that:
     • The Smith/Crapo bill was enacted in substantially similar form, and
     • It does not materially reduce deficits; in fact, it worsens them.
   • But the key quantitative part of the prediction—sustained annual deficits “roughly $2.5T” going forward—is about fiscal outcomes in the early‑to‑mid 2030s, well beyond the current date. Only projections exist so far; they can change, and the actual realized deficits for those years are unknown.

Because the decisive part of the prediction concerns deficit levels more than a decade out, and those years have not yet occurred, the prediction cannot yet be said to have “come true” or failed, even though current forecasts broadly align with Friedberg’s warning. Hence the appropriate status as of November 30, 2025 is inconclusive (too early).

• What was predicted? Friedberg described a hypothetical mechanism: if global investors began to seriously question the U.S. government’s willingness/ability to honor its 30‑year Treasury obligations, even a small increase in perceived default risk would push 30‑year yields up by ~1–2 percentage points into the 6–7% range, raising interest costs and annual deficits in a self‑reinforcing debt spiral.

• What actually happened (through Nov 30, 2025)?

  • 30‑year U.S. Treasury yields have been elevated relative to the 2010s but have not sustained anything close to the 6–7% range. In 2023–2024, they peaked around ~5% and have more commonly traded in the 4–5% band, driven largely by inflation, Fed policy, and term‑premium dynamics rather than an outright loss of faith in repayment of U.S. debt.
  • There has been no full‑blown market episode where investors broadly priced in a serious chance that the U.S. would not meet its long‑term Treasury obligations. While ratings downgrades (e.g., S&P in 2011, Fitch in 2023) and recurring debt‑ceiling standoffs have raised concerns about governance and fiscal trajectory, markets have not treated U.S. Treasuries as if default or non‑payment over 30 years were a live probability high enough to trigger a sudden jump to 6–7% yields.
  • The U.S. is experiencing rising interest costs and larger deficits, but these have been primarily associated with higher policy rates after the post‑COVID inflation spike and ongoing primary deficits, not with a discrete, confidence‑driven default scare on 30‑year Treasuries that matches the scenario Friedberg described.

• Why the result is ambiguous rather than right/wrong:

  • This is a conditional structural claim (“if investors start to doubt repayment over 30 years, then yields jump 1–2% into the 6–7% range and a debt spiral ensues”), not a dated forecast (“by year X, 30‑year yields will be 6–7%”).
  • The key condition—global investors meaningfully doubting the U.S.’s willingness/ability to honor 30‑year Treasuries—has not clearly occurred. Without that trigger, we don’t have real‑world data to test whether yields would in fact move specifically to 6–7% and whether this would mechanically produce the spiral in the way he outlined.
  • Economic theory suggests that a sharp upward repricing of default risk would raise long‑term yields and, all else equal, worsen debt dynamics, but the exact magnitude (1–2 percentage points, to 6–7%) and the inevitability of a runaway spiral are not empirically verifiable given current history.

Because the core scenario hasn’t actually materialized, we can’t decisively say Friedberg’s quantified causal chain is either confirmed or falsified. It remains a plausible but untested mechanism, so the appropriate classification is “ambiguous.”

The prediction is conditional on the outcome of the next U.S. federal election cycle after the episode date (May 17, 2025). In the U.S., regular federal general elections occur in even‑numbered years; after the November 2024 elections, the next regular federal cycle is the November 2026 midterm elections, which will decide control of Congress and therefore whether Democrats "come back into power." As of today (November 30, 2025), the 2026 elections have not yet occurred, so we cannot know whether Democrats regain the presidency and/or Congress, nor whether they subsequently "put a blockade" on large‑scale monetization of federal assets. Because the triggering condition and the predicted response both concern future events, there is not yet enough information to evaluate whether Friedberg’s forecast is correct.

The prediction is about whether over roughly the next 10–15 years new revenue from aggressively monetizing U.S. federal assets (land leases, resource extraction, etc.) will ramp up fast enough to fully make up for large projected federal budget deficits.

As of November 30, 2025, only months have passed since the May 2025 podcast—far short of the 10–15 year horizon. We therefore cannot yet observe the final outcome.

Current evidence does suggest that:

• Even with new Trump‑era policies (tariffs, the “One Big Beautiful Bill,” discussions of a sovereign wealth fund backed by public lands, increased leasing, possible land sales), federal deficits remain very large in official projections. (reddit.com)
• The Congressional Budget Office’s January and March 2025 outlooks project annual deficits around 5–6% of GDP and cumulative deficits of about $22 trillion from 2026–2035, with debt rising to roughly 118% of GDP by 2035, even under current law. (americanactionforum.org)
• Existing natural‑resource and public‑lands leasing revenues are in the tens of billions of dollars per year at most—tiny relative to multi‑trillion‑dollar annual budgets and near‑$2 trillion annual deficits. (reddit.com)

These data points so far are directionally consistent with Friedberg’s skepticism that asset monetization alone will quickly erase the deficit, but they are still just early‑period projections, not realized 10–15‑year outcomes. Major policy shifts, economic shocks, or changes in how assets are valued and sold could still occur before the horizon ends.

Because the stated time frame (10–15 years) has not elapsed, the prediction’s truth value cannot yet be definitively determined, so the status is “inconclusive” (too early).

The prediction’s horizon is “over the next several years … through roughly 2030,” so as of November 2025 it is too early to judge definitively.

Early evidence is mixed:

• Large-company restructurings and journalistic analyses show meaningful cuts to middle management layers and rising spans of control, with firms explicitly citing efficiency drives and AI tools as contributors to a “Great Flattening.” (wsj.com)
• Surveys and forecasts (e.g., Gartner) anticipate that AI will significantly reduce middle-management roles in a portion of organizations by the mid‑2020s, reinforcing the idea that managers are highly exposed to AI-enabled coordination and decision tools. (shrm.org)

However, aggregate labor data do not yet show a broad collapse in management jobs relative to other white‑collar roles. U.S. employment in management occupations is roughly flat-to-slightly up through 2024–2025 and is projected by BLS to grow about 7% from 2023–2033, while projected job losses from AI and automation are clearest in clerical/office support and some sales/customer-service roles rather than management per se. (bls.gov)

Because (1) the main evaluation date of ~2030 has not yet arrived, and (2) existing data show some AI-driven thinning of middle management but not a clear, systemwide pattern that managers are the first white-collar category to be materially reduced relative to individual contributors, the prediction cannot yet be labeled right or wrong.

Friedberg’s claim was that, because exits are weak, LPs would cut commitments and “you’re going to see a reduction in venture dollars”, i.e., a normalization of VC capital to the real pace of growth and innovation.

What happened to venture funding into startups?
Global VC investment in 2024 was already stabilizing well below the 2021 peak but modestly up from 2023 (about $368B vs. $346B).(practiceguides.chambers.com) In 2025, the value of PE/VC‑backed rounds in H1 rose ~25% year over year ($189.9B vs. $152.2B in H1 2024).(spglobal.com) KPMG’s Venture Pulse shows Q1 2025 at $128.4B and Q2 2025 at $101.05B,(kpmg.com) and then Q3 2025 climbing to about $120.7B, marking the fourth consecutive $100B+ quarter globally.(kpmg.com) Crunchbase‑based analyses concur that Q3 2025 funding (~$97B) was up ~38% year over year and that each of the last four quarters exceeded $90B, the strongest run since 2022.(news.crunchbase.com)(infotechlead.com)(eqvista.com) On this standard metric (capital deployed into startups), there has not been a broad reduction; funding volumes have rebounded and are running higher than in 2023–2024.

What happened to LP commitments / VC fundraising?
KPMG also reports that global VC fundraising by funds was very weak: only $48.8B raised by the end of Q2 2025, putting 2025 on track to fall below 2024’s eight‑year low of $187.3B.(mondaq.com) That is consistent with LPs being more cautious and reducing new commitments to the asset class, even as existing dry powder and capital concentration into AI megadeals keep deployed “venture dollars” high.(kpmg.com)(kpmg.com)

Why this is ambiguous rather than clearly right or wrong
If you interpret “reduction in venture dollars” as funds raised from LPs, then 2025 data so far supports his view: fundraising is materially down and likely to post a new cycle low, matching his story about LPs normalizing commitments.(mondaq.com) If you interpret it as capital invested into startups, the near‑term outcome has been the opposite: deal funding is up, with four strong quarters in a row, driven largely by AI and defense tech megadeals, and not yet “normalized” to a lower level.(news.crunchbase.com)(spglobal.com)(kpmg.com)

Because (a) his wording is ambiguous about whether he meant LP fundraising or capital deployed, (b) those two metrics have moved in opposite directions, and (c) only ~6–7 months of his “near‑to‑medium term” have elapsed, the prediction cannot be cleanly scored as fully right or fully wrong given the evidence available so far.

Key elements of Friedberg’s prediction have occurred.

1. Trump administration executive/regulatory action on foreign direct‑to‑consumer imports (including Amazon‑linked traffic)

   • On April 2, 2025, Trump signed Executive Order 14256, which eliminated the Section 321 de minimis exemption for all imports from China and Hong Kong effective May 2, 2025. This ended duty‑free treatment for low‑value parcels (<$800) from those locations and subjected them to tariffs and formal customs procedures. (en.wikipedia.org)
   • Reporting on the May 2 change explicitly notes that it hit Chinese platforms (Temu, Shein, AliExpress) and U.S. companies like Amazon, which had used de minimis to send low‑value goods directly to U.S. customers. (wired.com)
   • Earlier and subsequent tariff moves targeted the same loophole, with Reuters and others describing how shutting de minimis impacted Shein, Temu, and Amazon’s “Haul” discount channel, raising prices on their direct‑from‑China imports. (reuters.com)
   • Later in 2025, Trump signed an additional order ending the de minimis exemption globally as of August 29, 2025, so that all low‑value packages into the U.S. became subject to duties; coverage frames this as closing a major loophole that had enabled foreign e‑commerce sellers to ship goods directly to U.S. consumers duty‑free. (en.wikipedia.org)

   While these actions were not framed as being only about Amazon, they directly targeted the foreign low‑value import mechanism that Amazon’s cross‑border and third‑party sellers (e.g., Amazon Haul) had used, matching the substance of “acting on” abusive foreign direct‑to‑consumer import practices.

2. Congressional legislation to shut down or heavily restrict such foreign DTC imports

   • In mid‑2025, House Republicans advanced reconciliation legislation that proposed ending the Section 321 de minimis privilege worldwide on July 1, 2027 and adding sizable civil penalties for misuse, explicitly framed as cracking down on the de minimis import channel. (usfashionindustry.com)
   • Those provisions were enacted in Trump’s flagship reconciliation package, the One Big Beautiful Bill Act (OBBBA), which passed the House on May 22, 2025, the Senate on July 1, 2025, was agreed to again by the House on July 3, and was signed into law on July 4, 2025—all within two months of the May 2 podcast release. (en.wikipedia.org)
   • Trade‑law summaries confirm that OBBBA repeals the Section 321 de minimis exemption for commercial shipments from all countries effective July 1, 2027 and creates new civil penalties of up to $5,000 for a first violation and $10,000 for subsequent violations of Section 321, effective August 3, 2025. These analyses explicitly describe the aim as eliminating the de minimis privilege worldwide and punishing abuse of that import channel. (cmtradelaw.com)

   Taken together, this is precisely Congress “passing legislation” that shuts down and heavily restricts the key foreign direct‑to‑consumer import mechanism used by offshore sellers shipping directly to U.S. consumers.

3. Nuance: “foreign import of records”

   • A narrow reading of Friedberg’s off‑the‑cuff phrase about “foreign import of records” (vinyl, CDs, etc.) is not borne out literally. Due to the Berman Amendment, physical music media (including vinyl records) remain exempt from Trump’s tariffs and related measures as “informational materials,” and reporting in late 2025 notes that vinyl, CDs, and cassettes are specifically spared despite the broader de minimis crackdown. (nypost.com)
   • However, the normalized prediction you provided treats “foreign import of records” as shorthand for the broader foreign DTC loophole, and on that more general interpretation, both branches did exactly what he anticipated: an executive crackdown plus Congressional legislation to close the foreign low‑value import channel.

Bottom line: Within months of the May 2, 2025 episode, the Trump administration and Congress jointly executed a sweeping crackdown on the de minimis system that underpinned foreign direct‑to‑consumer imports (including Amazon‑related flows). That matches the core substance and timing of Friedberg’s prediction, even though the specific niche of imported vinyl “records” was not uniquely targeted.

As of 30 November 2025, the prediction has not materialized on any of its key points.

1. No evidence of Chinese 3 nm entering volume production in Q3 2025

   • Mainstream reporting still describes SMIC’s most advanced mass-produced node as 7 nm, constrained by U.S. export controls that block access to EUV lithography. Analyses note SMIC is researching 5 nm and 3 nm, but not yet producing them at scale, and remains at least several generations behind TSMC and Intel. (lemonde.fr)
   • Huawei/SMIC roadmaps consistently put true 3 nm chips in 2026: Huawei is reported to be working with SMIC on a 3 nm GAA design, with lab validation done and production-line adaptation underway, but the design is expected to be sent to SMIC and/or taped out in 2026, not manufactured in volume in 2025. (mobileworldlive.com)
   • Reports on China’s first domestic immersion DUV lithography tool say it targets roughly 28 nm initially and might be extended to 7 nm/5 nm via multipatterning, with integration into production lines only around 2027 and advanced-node mass production not expected before ~2030. That timeline is incompatible with broad 3 nm volume production starting in Q3 2025. (tomshardware.com)
     Together, these sources strongly indicate that while 3 nm R&D is underway, China did not have domestically-developed 3 nm technology in volume production by Q3 2025.

2. Conflicting niche articles don’t demonstrate Q3 2025 volume production

   • A ChinaCrunch piece from October 2025 claims SMIC has “successfully produces 3-nanometer (nm) chips domestically” and even says these are deployed in AI data centers and other applications. (chinacrunch.com) However, it does not tie this to a Q3 2025 start of volume production, does not quantify scale, and is not corroborated by Reuters, Nikkei, or other primary industry sources, which still describe SMIC as effectively capped at 7 nm mass production. (lemonde.fr)
   • Given the lack of independent confirmation and the continued depiction of SMIC as 7 nm–limited with yield problems at more advanced nodes, this article is better interpreted as highlighting experimental or very limited 3 nm capability, not the kind of high-yield volume production implied in the prediction.

3. No “significant share” of leading-edge capacity has shifted from Taiwan to mainland China

   • TSMC has been in high-volume 3 nm production since 2022; by Q3 2025, 3 nm processes contributed about 23% of its wafer revenue and 5 nm another 37%, meaning advanced nodes (3/5/7 nm) account for roughly three quarters of its sales. (stocks.observer-reporter.com)
   • TSMC’s roadmap shows 2 nm mass production ramping from late 2025 into 2026, with large 2 nm fabs being built in Hsinchu and Kaohsiung, reinforcing Taiwan as the core base of leading‑edge capacity. (roic.ai)
   • Meanwhile, coverage of China’s chip sector continues to describe SMIC as the only Chinese foundry at 7 nm and emphasizes its difficulty ramping that node due to export controls and domestic tool limitations—hardly consistent with capturing a significant share of global 3 nm/5 nm capacity. (marketscreener.com)
     There is no sign that a “significant share of leading-edge chip manufacturing capacity” has migrated from TSMC in Taiwan to mainland China by late 2025.

4. Why this is already judged wrong rather than ‘too early’
   The prediction had an early, time-bounded milestone (“go into [volume] production in Q3 of 2025”) that is already in the past and clearly unmet by any credible, independently corroborated evidence. Even optimistic Chinese and industry reports place genuine 3 nm manufacturing for Huawei/SMIC in 2026 and frame 2025 as an R&D and preparation year, not a volume‑production ramp. (mobileworldlive.com)
   Since this prerequisite step has effectively failed, and there has been no observable shift of leading-edge capacity from Taiwan to China, the overall forecast cannot plausibly come true in the form stated, even though 2026 is not yet over.

Conclusion: The available evidence shows no Chinese 3 nm volume production in Q3 2025 and no major relocation of leading-edge capacity from Taiwan to mainland China. On that basis, the prediction is best classified as wrong.

The prediction’s time horizon has not elapsed yet.

• The podcast was released on 5 April 2025, and the predictor said: “In the next 15 years…”, i.e., by around 2040.
• Today is 30 November 2025, so only ~8 months have passed since the prediction was made. There is no way to know yet whether by 2040 China’s added new electricity production capacity (in nameplate terms) will exceed 2× the 2025 total U.S. electricity production capacity.
• While current data and long‑term projections exist for China’s power-capacity build‑out and U.S. electricity production, those are forecasts, not realized facts, and can change substantially before 2040.

Because the deadline (2040) is still far in the future, the prediction cannot yet be judged as right or wrong. Therefore the correct classification is “inconclusive (too early)”.

The prediction is framed with a multi‑year horizon: that, as pending cases over officials’ use of private/ephemeral communications progress and in light of the Signalgate incident, U.S. law and policy will develop significantly clearer rules and consequences over the next few years.

Since the podcast (late March 2025), there have been concrete moves toward clearer policy on messaging apps and recordkeeping:

• The National Archives issued AC 23.2025 on May 2, 2025, explicitly addressing records created on third‑party messaging apps like Signal and WhatsApp. It reiterates that messages about agency business are federal records regardless of platform, sets out obligations when personal accounts are used (copy or forward to an official account within 20 days), details penalties for unlawful destruction, and warns that auto‑delete features can violate record‑keeping rules unless records are preserved first. It also clarifies what counts as a transitory vs non‑transitory message and how retention schedules apply. (archives.gov)
• Reporting indicates that, shortly after Signalgate, the White House instructed staff at the U.S. DOGE Service that work‑related messages sent on personal devices and on Signal must be archived, that auto‑delete must be disabled, and that the baseline rule is to preserve all work‑related communications regardless of format. (e-pluribusunum.org)
• Individual agencies have updated or reiterated policies requiring preservation of email, text, and instant messages that meet the definition of a federal record, prohibiting use of personal messaging accounts for agency business unless records are promptly forwarded to official systems. Recent examples include HHS records‑management policy and GSA’s 2025 email/IT data retention directive. (hhs.gov)
• In litigation stemming from Signalgate (American Oversight v. Hegseth), courts have ordered preservation of specific Signal chats and, in June 2025, ordered the acting archivist (Marco Rubio) to be notified about Signal chats at risk of deletion and to refer potential Federal Records Act violations to the attorney general for enforcement. That underscores the legal consequences side of the prediction. (americanoversight.org)

However, several factors keep this from being judged "right" or "wrong" yet:

• Much of the underlying legal framework (e.g., the content‑based definition of federal records and the extension of Capstone retention practices to electronic messages, including texts and third‑party apps) predates 2025. The 2025 developments largely clarify and reinforce existing rules rather than create a wholly new regime. (congress.gov)
• Ongoing controversies show that practice and enforcement are still unsettled. Watchdog groups continue to file complaints about officials using Signal with auto‑delete for official business even after Signalgate, arguing that this may violate the Federal Records Act and agency rules, which suggests that clarity and compliance are still being worked out. (theguardian.com)
• Most importantly, as of November 30, 2025 we are only about eight months into the "next few years" window the predictor invoked. There is still substantial time for additional legislation, regulations, and case law to emerge that could either strongly validate or undercut the claim that U.S. law and policy will become significantly clearer on which communications must be preserved versus can remain off the record.

Because the prediction is explicitly about developments over the next few years and we are still early in that period, with the legal/policy trajectory not yet fully settled, the status of the prediction is too early to call.

The new administration (Trump’s second term) began on January 20, 2025, so the first 100 days ran to about April 30, 2025.(en.wikipedia.org)

By that 100‑day mark, GDP data clearly reflected the impact of tariffs and spending cuts:

• Q1 2025 real GDP, covering January–March and released April 30, showed a –0.3% annualized contraction, the first since 2022. Multiple analyses attributed this to a record import surge as firms front‑loaded purchases ahead of Trump’s broad April tariff package, plus a sharp drop in federal spending.(cnbc.com)
• Detailed breakdowns noted imports up over 40% and federal government spending down about 5.1%, explicitly linking the decline to aggressive budget cuts and layoffs.(indilegalonline.com) A business‑economics review summed it up by saying Q1 GDP “highlight[ed] tariff impacts,” underscoring how the data revealed the directional effect of the new trade policy.(cbia.com)

Employment data in the same window captured the effects of federal cuts and policy uncertainty:

• February and March jobs reports showed moderating payroll growth (roughly 150k–220k per month) and unemployment edging around 4.1–4.2%, with commentary that mass federal layoffs from the new Department of Government Efficiency and tariff uncertainty were key headwinds, even as private‑sector hiring continued.(theguardian.com) This made the early labor‑market effects of the administration’s fiscal and trade stance visible.

Inflation data plus forward‑looking forecasts tied tariffs to the future price path:

• CPI reports for February and March 2025 showed inflation temporarily cooling (headline around 2.4–2.8% year‑over‑year), but economists repeatedly stressed that the announced April tariff package was not yet in the data and was expected to push prices higher later in 2025.(reuters.com) Major forecasters (e.g., Goldman Sachs) raised their projections for 2025 core PCE inflation to about 3.5% specifically because of the tariff program.(investopedia.com)

Markets and forecasters used these first‑100‑day data to reset expectations about growth, jobs, and inflation:

• As tariffs and spending cuts were rolled out and Q1 data arrived, Goldman Sachs and others repeatedly cut 2025 growth forecasts and lifted recession probabilities (to ~35–65%), explicitly citing the combined drag from higher tariffs, tighter fiscal policy, and weaker confidence.(fortune.com) A June 2025 NABE survey likewise showed consensus GDP forecasts marked down versus pre‑April and identified tariff impacts as the dominant downside risk to growth and employment.(nabe.com) Analysts also parsed Q1 GDP into “core” domestic demand versus tariff‑driven trade distortions to better project the path of real activity.(cbia.com)

Putting this together: within roughly the first 100 days, the observed trajectories of GDP (headline contraction with tariff‑driven trade distortion and reduced government spending), employment (slower gains with visible federal layoffs), and inflation (temporarily subdued but forecast to rise on tariffs) were already being used by economists and markets to model the directional impact of the administration’s spending cuts and tariff policy. That matches Friedberg’s claim that the first 100 days would be “pretty telling” for understanding those effects and for improving forward forecasts, so the prediction is best judged as right.

The prediction explicitly frames this as a story that will play out “over the next couple of years” after March 2025—a multi‑year horizon that, in the normalized version, runs roughly through 2027. The original quote in the episode is: “This is basically what I think is gonna end up being the story over the next couple of years, is the space race between effectively SpaceX's platform and technology versus China …” (podscripts.co)

As of today (2025‑11‑30):

• SpaceX’s Starship is still in a test‑flight phase but has become central to U.S. deep‑space and lunar plans, with multiple high‑profile test launches in 2025 and explicit ties to NASA’s Artemis program. (apnews.com)
• China’s Long March 9 is still an in‑development super‑heavy launcher, widely described as a Starship‑like, methalox, multi‑engine reusable heavy lifter, but its first test flight is generally targeted around 2030, well after 2027. (space.com)
• Broader space‑race coverage in 2024–2025 focuses on a U.S.–China competition in space (lunar and Mars ambitions), but also heavily features NASA’s Artemis/SLS, Blue Origin’s New Glenn, and other U.S. commercial players. NASA has even reopened parts of the lunar‑lander competition beyond SpaceX, which dilutes any simple “Starship vs Long March 9” framing as the dominant narrative. (livescience.com)

Because (1) the stated time window (through ~2027) has not yet elapsed and (2) narrative dominance is inherently subjective and can’t be settled this early in that window, there isn’t enough evidence yet to say the prediction is clearly right or clearly wrong. Hence the result is inconclusive (too early).

The prediction’s horizon is “the next couple of years” from the podcast date in March 2025, i.e., roughly into 2026–2027, so that full window has not elapsed yet.

On the facts:

• Starship has made multiple orbital‑class test flights in 2025 (Flights 7–11) and is clearly positioned as the U.S. flagship for future super‑heavy, fully reusable launch, but it remains in a test/early demonstration phase, not yet an established operational workhorse or the sole focus of U.S. strategic launch. (en.wikipedia.org)
• Long March 9 has been repeatedly redesigned into a Starship‑like, fully reusable methane/LOX super‑heavy rocket and is explicitly described as a Starship‑class competitor, but it is still on the drawing board, with a first flight currently targeted around 2033, well after the mid‑2020s. (en.wikipedia.org)
• China’s near‑term crewed‑lunar strategy for the 2020s instead centers on Long March 10 and existing Long March rockets, while Long March 9 is framed more as an infrastructure builder for 2030s lunar bases. (en.wikipedia.org)
• The current strategic launch landscape in the mid‑2020s is multi‑polar: besides SpaceX’s Falcon 9/Heavy and Starship tests, the U.S. has ULA’s Vulcan Centaur and Blue Origin’s operational New Glenn (with successful launches and an NSSL role), and Europe’s Ariane 6 has entered commercial service; these systems all feature in discussions of launch competition. (en.wikipedia.org)
• Geopolitical commentary broadly describes a new space race between the U.S. and China (sometimes including Russia), focused on lunar landings, bases, and even nuclear propulsion, rather than specifically and exclusively on a Starship‑vs‑Long‑March‑9 duel in the mid‑2020s. (houstonchronicle.com)

Because:

1. we are only partway through the “next couple of years” window implied by the quote, and
2. Long March 9 has not flown and is not yet central to current operational competition (its role is still prospective), we cannot yet definitively say whether Starship vs. Long March 9 will emerge as the primary strategic launch rivalry over the full horizon the predictor specified. Hence the status is inconclusive (too early) rather than clearly right or wrong.

As of November 30, 2025, both of the predicted missions are still in the future:

• Multiple sources describe Blue Ghost Mission 2 as scheduled to launch in 2026 but do not give a firm launch date or indicate that it has already flown. (en.wikipedia.org)
• Firefly and NASA press-related materials describe Blue Ghost Mission 3 as a mission to the Gruithuisen Domes slated for 2028, again with no indication that it has occurred yet (which would be impossible given the date) or that it has definitively slipped beyond 2028. (news.flyjets.com)

The prediction is about whether Mission 2 will actually fly in 2026 and Mission 3 in 2028. Since we are still in 2025, neither year has arrived, so we cannot yet judge success or failure of those schedule claims. The current information only shows intent and planning, not realized outcomes.

Because the relevant years have not yet passed, the correct status for this prediction is “inconclusive (too early)”.

As of November 30, 2025, no fertility clinic is routinely offering IVF using human egg cells derived from a patient’s somatic cells (e.g., skin cells) via in‑vitro gametogenesis (IVG) in clinical practice. Research groups have generated functional egg and sperm cells from somatic or pluripotent cells in mice and other animals, and human IVG has advanced in vitro (e.g., creating primordial germ cell–like cells and early gamete precursors), but no approved, routine clinical IVF protocol exists yet for producing fully functional human eggs from skin cells and using them in patients.

However, the normalized prediction is explicitly about what will be routine "by roughly the late 2020s". The current date is still only 2025, so the time window for the prediction (late 2020s ≈ 2027–2029) has not arrived. Because we cannot yet observe whether this will become routine by that future period, the correctness of the prediction cannot be determined at this time.

Therefore, the appropriate status is: it is too early to tell whether his forecast about late‑2020s routine clinic use will prove right or wrong.

As of 30 November 2025, the time window specified by the prediction (“in the next couple of years” / “by the late 2020s”) has not yet elapsed, so it is too early to know whether IVF will in fact become significantly simpler, less invasive, and much more widely available than in early 2025.

Current scientific progress supports that we are still in an early, experimental phase:

• In 2025, scientists for the first time turned human skin cells into egg-like cells that could be fertilised, a key in‑vitro gametogenesis (IVG) milestone. However, the embryos showed many abnormalities, and researchers stress this is only proof‑of‑concept and estimate that such lab‑generated eggs are at least a decade away from being widely available as a treatment. (tribune.com.pk)
• The UK fertility regulator similarly concluded in 2025 that lab‑grown eggs and sperm are on the brink of viability as a research technology, but clinical use is still expected to be up to a decade away, not something available broadly in the mid‑2020s. (theguardian.com)

Because (1) the forecast’s horizon extends into the late 2020s, and (2) we are only partway through that period, with the decisive technologies (like lab‑generated eggs) not yet clinically deployed, the correct judgment as of now is inconclusive rather than right or wrong.

It’s too early to know whether this prediction is correct or not.

The claim is that, under current law and contribution/benefit patterns, the U.S. Social Security Trust Fund will become insolvent around 2032 ("going to go bankrupt in 2032"). Insolvency here means the trust fund reserves are depleted and the system can no longer pay full scheduled benefits, not that payments stop entirely.

As of November 30, 2025, that future date has not yet arrived, so the trust fund has not become insolvent and no definitive judgment about 2032 can be made.

For context, recent official projections are in the same general time frame but do not exactly match 2032:

• The 2024 Social Security Trustees Report projects that the combined OASDI trust funds will be depleted in the mid‑2030s under current law (with the exact year depending on the report’s assumptions and whether you look at OASI alone or combined OASDI). After depletion, incoming payroll taxes would still cover roughly three‑quarters of scheduled benefits, but not 100%.
• Other analyses (e.g., from the Congressional Budget Office) have at various times projected depletion dates around the early‑to‑mid 2030s as well.

Because (1) the predicted year 2032 is still in the future and (2) projected depletion dates can shift with each new report and policy/economic change, this prediction cannot yet be judged as right or wrong.

Available evidence shows that discussion of replacing the U.S. income‑tax system with a consumption‑tax system (e.g., national sales tax, VAT‑style proposals) remained a recurring but niche theme in 2025, not a clearly more prominent or central topic than in the years immediately before.

Baseline (before 2025):

• In early 2023, the FairTax Act of 2023 (H.R. 25) to replace federal income, payroll, estate, and gift taxes with a national sales tax drew high‑level political fire: President Biden publicly attacked it as a national sales tax that would raise middle‑class costs, and Senate Democrats introduced a resolution condemning it, both widely covered in mainstream outlets. (congress.gov)
• Media and policy analysis in 2023–24 already framed FairTax and related national sales‑tax ideas as significant Republican proposals, with fact‑checking and think‑tank pieces explaining their mechanics and distributional effects. (fairtax.org)
• By 2024, Project 2025 was being heavily covered as a far‑right blueprint, explicitly highlighting its long‑term goal of replacing income and corporate taxes with a national consumption tax (national sales tax, VAT‑type options, flat consumption tax). Major outlets like the Guardian, Forbes, and advocacy groups dissected these consumption‑tax elements as a key part of the plan. (forbes.com)

This means that before 2025 there was already visible, national‑level political and media debate about moving from income to consumption taxation.

During 2025:

• The main enacted tax legislation in 2025—the One Big Beautiful Bill Act, signed July 4, 2025—extends and tweaks income‑tax provisions (extending 2017 individual rate cuts, altering SALT caps, creating deductions for overtime, tips, etc.) but does not implement or seriously advance a federal consumption tax. The structure of federal revenue remains income‑tax‑centric. (en.wikipedia.org)
• Project 2025’s consumption‑tax vision continues to be discussed in 2025 analysis pieces (e.g., Forbes Advisor; Center for American Progress), emphasizing that the blueprint envisions eventually moving from an income tax to a consumption tax and modeling how such a shift would affect middle‑class vs. wealthy households. But these are largely continuations of the 2024 debate over Project 2025, not a qualitative step‑change in salience. (forbes.com)
• Congress has ongoing background work and CRS material on consumption taxes and the reintroduced FairTax‑style bill, but these remain at the level of "one option for fundamental tax reform" and "an alternative model" under consideration, with no major new legislative push in 2025 comparable to, say, the 2023 FairTax Act media moment. (congress.gov)

Putting this together:

• The topic exists in 2025 and continues to be discussed, especially in the context of Project 2025 and academic/CRS analyses.
• However, the visibility and intensity of discussion were already high in 2023–24, with presidential speeches, Senate resolutions, and substantial media coverage centered on FairTax and national‑sales‑tax proposals. 2025 coverage largely extends those conversations rather than showing a clearly noticeable increase or making the shift to consumption taxation one of the dominant economic/political topics of the year (especially relative to tariffs, income‑tax cuts in the One Big Beautiful Bill, inflation, etc.).

Because the prediction specifically claimed we would "hear about it a lot more this year" and that the shift from income to consumption taxation would become a prominent topic in 2025 U.S. debate, and the evidence instead shows continued but not clearly elevated or central discussion, the prediction is best judged as wrong, not merely undecidable.

The prediction was framed over a 1–2 year horizon from March 2025 (i.e., roughly until March 2027), so as of November 30, 2025 the full timeframe has not elapsed.

Evidence so far is mixed:

• Hyperscalers are indeed massively expanding and bundling AI capacity. AWS and Google Cloud have rolled out successive generations of custom AI chips (Trainium / Inferentia, TPU v5e/v5p, v6 Trillium, v7 Ironwood) and large-scale GPU clusters, explicitly marketed as cheaper, integrated options for AI training and inference. (michaelbommarito.com)
• GPU compute prices are already under downward pressure. Analyses of H100 rental markets in 2025 report a substantial increase in supply and intense competition, including a ~44% AWS price cut on H100 instances and broad price reductions across many clouds, with H100 rental rates drifting down across the market. (intuitionlabs.ai)
• However, demand for specialized GPU clouds like CoreWeave has not “significantly reduced” so far; it remains very strong. CoreWeave has continued rapid expansion in 2025: it signed an ~US$12B, five‑year cloud deal with OpenAI, grew to 32 data centers with ~250,000 GPUs, became the first to offer Nvidia GB200 NVL72 (and later Blackwell Ultra) in the cloud, and completed a large IPO despite leverage and governance concerns. (en.wikipedia.org) Analyst commentary in mid‑2025 still highlights strong projected revenue growth into 2026–27 rather than collapsing demand. (barrons.com)
• Other “neo‑cloud” GPU providers also show robust demand. For example, Nebius (another specialized AI cloud) reported several‑hundred‑percent revenue growth in 2025 and signed multi‑billion‑dollar, multi‑year infrastructure deals with major customers including Meta and Microsoft, noting demand was strong enough that one deal had to be capped by available capacity. (reuters.com)

So far, the observable market aligns with the pricing‑pressure part of the thesis (falling GPU rental prices as capacity ramps), and with hyperscalers increasingly bundling AI compute into their platforms. But the “significantly reduced demand for third‑party GPU clouds” component has not materialized yet; if anything, those providers are still rapidly scaling and signing large contracts. Because the prediction’s endpoint is still ~16 months away and current data do not decisively confirm or refute a future demand drop by 2027, the outcome must be judged inconclusive (too early) rather than clearly right or wrong at this time.

Evidence to date lines up closely with Friedberg’s forecast on both the debt‑refinancing and stock‑market‑tolerance components.

1. Context: Trump 2.0 and Bessent in key roles. Donald Trump began his second, non‑consecutive term as president on January 20, 2025, and Scott Bessent was confirmed as Treasury secretary later that month, matching Friedberg’s assumption that “Bessent and others” would hold key economic posts in a new Trump administration. (en.wikipedia.org)

2. Massive near‑term refinancing need (~$10–11T). Treasury’s Office of Debt Management reported in mid‑2025 that about 31.4% of the roughly $36 trillion federal debt would mature within the following 12 months, meaning about $11 trillion in U.S. debt securities must be refinanced—almost exactly the “$10 trillion coming due in the next 12 months” Friedberg referenced. (foxbusiness.com) This refinancing window overlaps the 12‑month period he was talking about (roughly March 2025–March 2026).

3. Administration actively trying to push rates down around that refinancing. Early in the term, Bessent said the Trump administration’s focus was on keeping longer‑term Treasury yields (especially the 10‑year) low, rather than directly browbeating the Fed, and confirmed that “the president wants lower rates,” aiming to get borrowing costs down via deregulation, spending cuts, and energy policy. (cnbc.com) As the year progressed and the refinancing wave loomed, Trump repeatedly demanded deep Fed cuts, arguing publicly that the policy rate was at least 3 percentage points too high and that each percentage point was costing the government roughly $360 billion per year in refinancing costs—explicitly tying his push for lower rates to the cost of rolling the debt. (reuters.com) Bessent later stated that the Fed’s rate was “significantly higher than necessary” and called for a series of cuts totaling about 150–175 basis points, while also emphasizing that rising tariff revenues should be used to reduce debt and that rate cuts would ease pressure on rate‑sensitive sectors. (upi.com) Together with Treasury’s own acknowledgment that roughly $11T must be refinanced over a one‑year horizon, this amounts to exactly the sort of strategy Friedberg described: using political and policy pressure to push interest rates down during a huge refinancing window.

4. More willingness than in term one to tolerate stock‑market weakness. In his first term, Trump frequently treated record stock prices as a core metric of success and regularly highlighted market gains. (en.wikipedia.org) In 2025, by contrast, he pressed ahead with his sweeping "Liberation Day" tariff program even as it triggered the largest U.S. stock‑market crash since the pandemic and the worst first‑100‑days market performance for any president since Gerald Ford, with the S&P 500 down over 7% and trillions in equity value temporarily wiped out. (en.wikipedia.org) Investment research notes from March 2025 observed that markets were starting to question the old “Trump put,” explicitly commenting that the new administration appeared willing to tolerate a “temporary disturbance” in economic activity and equity prices to pursue its policy agenda. (ubs.com) During sell‑offs, Trump himself said that while he didn’t want stocks down, “sometimes you have to take medicine to fix something,” and he openly acknowledged the risk of a recession, signaling acceptance of short‑term market and economic pain. (reddit.com) This behavior is widely described by market commentators as a shift from his first term, when he was more visibly reactive to market declines.

5. Linking both elements together. Commentary on the 2025 crash and tariff shock notes that some investors even speculated Trump might be willing to endure market turmoil in order to force or justify easier monetary policy, though he denied doing it "on purpose." (reuters.com) At the same time, reputable financial and policy outlets tied Trump’s and Bessent’s aggressive calls for lower rates directly to the government’s need to refinance an ~$11 trillion debt wall and reduce exploding interest costs. (foxbusiness.com) This is closely aligned with Friedberg’s thesis that Trump 2.0 would be more willing than before to look past short‑term stock‑market "rumblings" in order to get rates down and refinance the debt on more favorable terms.

Given the documented combination of (a) an explicit, sustained push from Trump and Bessent to lower interest rates precisely as a massive ~$10–11T refinancing wave hits, and (b) a noticeably greater tolerance for equity‑market drawdowns compared with his first term, Friedberg’s prediction has, by late 2025—well within his 12‑month horizon—substantially come true.

The prediction’s own timeline runs “in the next couple of years” from March 8, 2025 (roughly out to 2027–2028), so as of November 30, 2025, we are far from the end of the forecast window. It is therefore too early to say definitively whether it will prove right or wrong.

Evidence so far is mixed and mostly against the idea that NATO is broadly seen as “not necessary”:

• US and NATO elites are doubling down on the alliance, not deemphasizing it. The 2024 Washington NATO summit and the 2025 The Hague summit reaffirmed NATO as central to Euro‑Atlantic security and produced a high‑profile pledge to raise defense spending to 5% of GDP by 2035, framed explicitly as a response to enduring Russian and other threats, with an “ironclad” commitment to Article 5. (en.wikipedia.org) Senior US and NATO figures (e.g., Nancy Pelosi, Secretary‑General Mark Rutte) publicly stress that the US “needs NATO too” and that this is “not the time to go it alone,” describing NATO as irreplaceable for collective security. (politico.com)

• US public opinion still broadly sees NATO as beneficial and necessary. An April 2025 Pew survey finds 60% of Americans view NATO favorably and 66% say the US benefits from membership, with particularly strong support among Democrats. (pewresearch.org) A May–June 2025 Chicago Council/Ipsos poll reports that 74% of Americans favor maintaining or increasing the US commitment to NATO and 57% say NATO makes the US safer. (globalaffairs.org) A large YouGov survey likewise finds about 71% of Americans support NATO and 66% back its expansion. (today.yougov.com)

• There is, however, a notable segment moving in the direction Friedberg describes. Pew and Chicago Council data show Republican and especially conservative‑Republican skepticism about NATO has grown: only about half of Republicans say the US benefits from membership, and rising shares say the US does not benefit, while figures like Donald Trump and Elon Musk have openly floated reducing or even ending US participation. (en.wikipedia.org) That’s a real elite‑and‑base shift, but it is still counterbalanced by strong overall support and increased alliance investment.

Given that (a) the prediction horizon runs to ~2028 and (b) current evidence shows strong but increasingly polarized support rather than a broad consensus that NATO is “less necessary,” the correct assessment as of late 2025 is that it is too early to determine whether the prediction will ultimately be right or wrong.

As of November 30, 2025, Trump’s “gold card” / Gold Card visa concept is still in flux and has not been operating long enough to know its total lifetime demand. Trump initially announced a $5 million “gold card” investor visa in February 2025, talking about selling up to a million cards, but for months there was no formal application process or clear legal authority, and immigration-law analyses emphasized that the program effectively did not yet exist in U.S. immigration law. (reuters.com) In September 2025, he signed an executive order launching a Trump Gold Card priced at $1 million (or $2 million via corporate sponsorship), while moving the $5 million level to a separate “Platinum Card” product that is not yet operational and would require congressional changes to immigration and tax law. (forbes.com) Various reports mention tens of thousands of people registering interest and administration claims that 1,000 gold cards were “sold” in a day, but investigative coverage and practitioner briefings note that, through at least mid‑2025, there was no functioning application channel, no way to pay, and thus no verified issuance of actual cards. (visaverge.com) Industry experts widely doubt the administration’s projection of a million buyers and instead project demand in the low thousands or even under 1,000 total cards, but these are forward‑looking opinions rather than observed outcomes. (uscis.techmis.com) Because the program’s design, price point, and legal status are still unsettled, and there is no reliable data yet on actual long‑term uptake, Friedberg’s prediction that total lifetime demand will be capped around 10,000 buyers and never reach one million cannot be confirmed or disproven at this time.

The prediction is about the number of buyers in the first full year after the program actually launches.

Timeline and status:

• Trump announced the idea of a $5M “gold card” / investor green card replacement for EB‑5 in late February 2025, with further promotion and a physical card shown to reporters in early April 2025. (investing.com)
• Reporting through mid‑2025 (e.g., Washington Post, Henley & Partners) notes that the program faced serious legal obstacles, that no new visa category had been created by Congress, and that any claims of cards already being “sold” were not backed by official statistics and treated skeptically. (washingtonpost.com)
• On September 19, 2025, Trump signed an executive order directing the Secretary of Commerce to establish the Gold Card program, but this order still left implementation details to agencies. (whitehouse.gov)
• As of November 21, 2025, an immigration law update based on USCIS materials says that applications are not yet being accepted, with agencies only previewing forms and rules and targeting a future launch (described as a December start), and it provides only an internal estimate of roughly 1,000 applicants per year going forward. (greencardmaker.com)

Crucially, there is still no transparent, official data on how many Gold Cards have actually been purchased or issued, and the program’s first operational year has either not begun or is only just beginning as of late November 2025. That means we cannot yet observe the realized number of “buyers in the first year after launch,” nor can we reliably infer it from public sources.

Because the relevant one‑year period has not elapsed and no authoritative issuance statistics are available, whether the eventual buyer count will fall in the 1,000–2,500 range is too early to determine.

As of November 30, 2025, the full conditional scenario Friedberg described has not clearly played out, and there is no explicit time horizon in his prediction.

Precondition – broader retail access: In 2025 the SEC did take meaningful steps toward expanding ordinary investors’ access to private markets. Most notably, it stopped enforcing a staff policy that had capped registered closed‑end funds’ investments in private funds at 15% unless sales were limited to accredited investors with a $25,000 minimum. This change now allows such funds to invest more heavily in private funds (including private‑equity and venture strategies) while being sold to non‑accredited retail investors, effectively broadening indirect retail access to private markets.

• SEC staff and industry commentary describe this as providing greater access for non‑accredited investors to private equity and private credit through retail closed‑end funds and related vehicles. (wssllp.com)

However, this is primarily fund‑mediated access to diversified private funds, not unfettered direct investing by non‑accredited individuals into individual private startup equities at large scale. So even the precondition (“can broadly invest in private startup equity”) is only partially met, and only recently.

Downstream effects (losses and backlash): I find no evidence of a major, widely reported wave of non‑accredited retail investors losing “substantial” amounts of money specifically because of this new access regime to private startup/PE style equity. Commentary and SEC advisory work emphasize potential risks and the need for safeguards, not an already‑materialized retail wipe‑out. (barrons.com)

Elizabeth Warren’s response: Senator Elizabeth Warren has continued her long‑standing criticism of private equity and private markets – e.g., letters and legislation targeting private‑equity buyouts (Walgreens/Sycamore), the Stop Wall Street Looting Act, and concerns about private equity and private credit in retirement accounts. (warren.senate.gov) But these positions predate and are not clearly a new political backlash tied to a fresh wave of losses by newly empowered non‑accredited startup investors, nor have they yet resulted in enacted rules that specifically roll back the 2025 retail‑access expansions.

Given:

• the enabling policy changes are relatively recent and only partially match the “broad” direct startup‑equity access condition;
• there is no documented large‑scale loss event among that new investor cohort; and
• Warren’s ongoing push for stricter private‑equity regulation is part of a long‑running agenda rather than a clearly new, loss‑triggered backlash to these 2025 access changes,

there isn’t enough evidence yet to say that Friedberg’s full causal chain has either definitively occurred or definitively failed. It remains too early to judge, so the prediction is best classified as inconclusive.

The prediction explicitly concerns a “post‑AI era” over the coming decades, in which AI allegedly makes routine things like:

• Deep mining toward the Earth’s interior ("500 miles down") for rare earth minerals
• Large‑scale space projects such as colonizing the Moon

As of November 30, 2025, none of this has occurred:

• Current mining operations are limited to relatively shallow depths in the Earth’s crust; there is no technology or active industrial practice approaching hundreds of miles deep extraction for rare earths or anything else.
• Human activity on the Moon is limited to robotic missions and planning for future crewed missions (e.g., NASA’s Artemis program and various commercial lunar initiatives). There is no permanent lunar colony or routine colonization activity in operation yet.
• AI has advanced rapidly and is being used in engineering, logistics, and scientific research, but it has not yet demonstrably enabled the kind of massive, routine, planetary‑scale and space‑colonization projects described.

Because the prediction’s implied time horizon is “the post‑AI era” over the coming decades, and we are only in 2025, it is too early to determine whether AI will in fact unlock the level of routine deep‑Earth mining and Moon colonization described. The relevant future period has not arrived, and the prediction is not framed to be testable this soon.

Therefore, the status of the prediction is inconclusive (too early to tell) given the current date.

The asteroid in question is 2024 YR4, an Apollo‑type near‑Earth object whose orbit crosses Earth’s and which has a possible close approach on 22 December 2032.

• Around 19–20 February 2025, NASA’s Center for Near‑Earth Object Studies (via Sentry) and ESA indeed reported that 2024 YR4 had an impact probability of about 1.5% for 2032, with an estimated size roughly 40–90 m (≈130–300 ft) across. Multiple outlets summarized this as a 1.5% chance of impact after a peak of about 3.1%, matching Friedberg’s “~1.5%” description tied explicitly to the then‑current trajectory estimates. (snopes.com)

• Within days, new observations rapidly drove the modeled impact probability down: Sky at Night and other sources note that the chance fell from 3.1% to 1.5% on 19 Feb, to 0.28% on 20 Feb, and then to about 0.005% by 23 Feb 2025. (skyatnightmagazine.com) Subsequently, NASA and ESA further reduced the risk to roughly 0.001–0.004%, effectively zero, and the object was downgraded to Torino Scale 0. (apnews.com)

• Energy‑yield estimates from NASA‑linked analyses and expert commentary put a 40–90 m stony asteroid like 2024 YR4 in the rough range of ~7–10 megatons of TNT if it actually hit Earth—enough for severe city‑scale damage but not a global catastrophe. (en.wikipedia.org) Friedberg’s “up to ~20 megatons if it’s near 300 feet” is a back‑of‑the‑envelope extrapolation from the upper end of the size range and cannot be directly validated or falsified without an actual impact.

• Crucially, the core of his statement is a probabilistic forecast about whether the asteroid will hit Earth in 2032 and, if so, the chance of human casualties. As of 30 November 2025, 2032 is still in the future, and 2024 YR4 has obviously not impacted Earth yet. The updated orbital solutions now imply an extremely small impact probability, but whether the true physical outcome in 2032 is “no impact” (which would mean the real ex ante probability was effectively near zero) cannot be definitively judged until after that date.

Because the underlying event (any 2032 impact and associated loss of life) has not yet occurred, the correctness of Friedberg’s forecasted 1.5% impact chance and ~15% conditional fatality chance cannot be empirically evaluated. His descriptive claim about the then‑current NASA estimate (~1.5%) was accurate, but the predictive content about the actual 2032 outcome remains untestable at this time. Hence the status is inconclusive (too early to tell).

NASA’s own Planetary Defense blog update on Feb. 20, 2025 reports that observations overnight on Feb. 19–20 reduced asteroid 2024 YR4’s chance of impacting Earth on Dec. 22, 2032 to 0.28% (about 1 in 360), i.e., slightly under one‑third of one percent. Independent coverage of the same NASA update likewise cites a new impact probability of 0.28–0.3%, often rounded to 0.3%, and describes it as “about 0.3%.” This is consistent with Friedberg’s wording that NASA had “dropped the probability … to about a one third of 1%,” so his characterization of the updated February 2025 NASA assessment is essentially accurate, even though NASA later reduced the probability much further (to ~0.004%) in subsequent days.

It is too early to judge this prediction as right or wrong.

Key parts of Friedberg’s claim:

1. It is explicitly conditional and long‑term:

   • “If” the EU and U.S. adopt a techno‑pessimistic, highly regulatory stance on AI and
   • “over the coming years” China’s GDP will scale while U.S. GDP stagnates relative to China.

   As of November 30, 2025, less than one year has passed since the February 15, 2025 podcast. That is far too short to evaluate multi‑year structural GDP divergence.

2. Policy condition is not clearly resolved yet:

   • The EU has indeed been moving toward heavy AI regulation via the EU AI Act, often cited as one of the world’s most restrictive AI regulatory frameworks.
   • The U.S. has taken a more fragmented approach: an AI Executive Order, sectoral proposals, and some state‑level rules, but not yet a comprehensive, clearly “techno‑pessimistic” regime. Policy direction is still evolving and contested.
   • China has been promoting rapid AI deployment within a state‑controlled framework, combining industrial policy support with content and safety regulations; whether this is best characterized as "embracing" vs. "constraining" AI is itself debated.

   Because the exact scenario he described (strongly pessimistic, highly regulatory U.S. and EU versus a clearly more liberal, growth‑oriented China) is still in flux, the antecedent of the “if…then…” prediction is not firmly in place yet.

3. GDP outcomes have not had time to manifest in the structural way implied:

   • Real GDP levels and growth for 2024–2025 for the U.S. and China can be estimated or projected by IMF/World Bank/OECD data, but a single year or two of growth data cannot confirm a multi‑year structural stagnation of U.S. GDP relative to China driven specifically by AI regulatory stance.
   • Even if early data showed China growing somewhat faster or slower than the U.S., attributing that to AI regulatory philosophy (rather than real estate cycles, demographics, trade, fiscal policy, etc.) would be speculative at this stage.

Because:

• the time horizon (“over the coming years”) has not elapsed,
• the policy precondition is not fully realized or stable, and
• GDP trends and causal attribution cannot yet be robustly assessed,

this prediction cannot be declared right or wrong at this time and is best classified as inconclusive.

Key parts of Friedberg’s scenario have already occurred, but the decisive "tens of billions" in new tariff‑driven farm transfers are not yet clearly in place, and Trump’s second term is still in its first year.

• A future Trump administration did in fact impose a new round of broad tariffs on China in 2025, including a 10% baseline tariff on almost all imports and a higher “reciprocal” rate on China of about 34%, on top of earlier 2025 measures that doubled tariffs on Chinese goods from 10% to 20%. (en.wikipedia.org)
• China responded exactly as predicted in direction: it imposed additional 10–15% retaliatory tariffs on a wide range of U.S. agricultural and food products (covering hundreds of tariff lines) and also suspended soybean import licenses for several U.S. firms and halted some U.S. log imports, clearly targeting U.S. agriculture. (fas.usda.gov)
• However, while there are substantial government payments to farmers in 2025, most are from pre‑existing or broadly framed programs (e.g., a $10 billion Emergency Commodity Assistance Program for low prices, of which over $8 billion has been paid, plus large disaster payments projected at about $35.7 billion) rather than a clearly identified, new, tariff‑specific bailout on the scale of the 2018–19 trade‑war programs. (reuters.com)
• A new trade‑related bailout in the $10–15 billion range has been reported as planned, and over $3 billion in CCC‑funded aid tied to trade disputes is being released, but reporting as of late October–November 2025 treats these packages as pending or partial; analysts describe tens of billions in tariff‑funded farm aid as a likely outcome if the trade war continues, not as an accomplished fact. (reuters.com)

Because (1) the tariff and Chinese‑retaliation components of the prediction are already borne out, but (2) the scale and explicit tariff‑linkage of new farm transfer payments have not yet clearly reached “tens of billions” attributable to the 2025 China episode, and (3) Trump’s term runs through January 2029, it is too early to say definitively whether the full prediction has come true. Hence the result is classified as inconclusive rather than right or wrong at this time.

In 2025 the U.S. did enact exactly the kind of new, broad tariff rounds Friedberg was talking about: Trump layered a 10% “fentanyl” tariff on all Chinese imports, added sweeping reciprocal tariffs of 10–50% on many countries, and imposed 25% duties on Canada, Mexico, and others under various executive orders. (reuters.com) China initially retaliated with steep tariffs but then agreed to a series of negotiated truces: a May 2025 Geneva deal slashed reciprocal rates from well over 100% to 10% for both sides, followed by an October accord that locked in lower tariffs for a year and bundled Chinese commitments on fentanyl control and resumed U.S. farm purchases, i.e., clear partial concessions within a negotiated settlement rather than sustained maximal retaliation. (china-briefing.com) Canada first imposed broad 25% counter-tariffs, then in August 2025 dropped most of them on USMCA‑qualifying goods to “match” U.S. exemptions, a move widely described as easing or retreating from its hard line, while sectoral tariffs on U.S. steel, aluminum, and autos remained – again, a partial concession embedded in ongoing talks rather than indefinite full retaliation. (cnbc.com) Mexico, similarly, secured exemptions for USMCA‑compliant exports from the global tariff hike and later obtained a 90‑day reprieve from additional planned U.S. tariffs to negotiate a broader arrangement, instead of implementing or sustaining maximum counter‑tariffs of its own. (reuters.com) The EU suspended planned 25% countermeasures, then agreed to a July–August 2025 framework where it will eliminate tariffs on U.S. industrial goods and grant preferential access for many U.S. farm and seafood exports, while accepting a 15% U.S. tariff on most EU goods and pausing its retaliation for at least six months – a textbook negotiated compromise with significant European concessions. (theguardian.com) Some disputes, notably with India, remain in flux, but even there officials on both sides now openly expect a tariff truce and trade deal that would adjust the punitive U.S. rates, not lock in India’s maximum retaliatory posture. (reuters.com) Taken together, the major trading partners targeted in this new tariff round have indeed moved toward partial concessions and negotiated settlements rather than maintaining maximal retaliatory tariffs indefinitely, matching Friedberg’s prediction in substance.

As of November 30, 2025, the U.S. Social Security trust funds have not yet become insolvent, and the prediction is about a status change expected around 2033, which is still in the future.

Current projections from the Social Security Trustees (2024 report) estimate that the combined Old-Age and Survivors Insurance (OASI) and Disability Insurance (DI) trust funds will be depleted in the early–mid 2030s (often cited as around 2033–2035), after which incoming payroll taxes would cover only about three-quarters of scheduled benefits, implying either automatic benefit cuts or the need for additional financing. These projections are broadly consistent with the timeframe mentioned in the prediction, but they are forecasts, not events that have already occurred.

Because insolvency has not yet occurred and the relevant year (≈2033) is still ahead of us, there is not enough information to say the prediction is right or wrong at this time. Therefore the correct status is: inconclusive (too early to tell).

Friedberg forecast three linked steps: (1) an aggressive wave of DOGE/Trump spending cuts within the next few months, (2) court fights that clarify what the executive can and cannot cut unilaterally, and (3) a shift toward pushing cuts through Congress.

1. Whirlwind of cutting: After Trump’s January 20, 2025 order creating DOGE, the administration rapidly rolled out mass layoffs, workforce‑optimization orders, contract cancellations and grant terminations across many agencies in February–April 2025, including large reductions at HHS, Education, HUD, VA, IRS and others. (en.wikipedia.org) That fits a fast, aggressive attempt to cut spending of exactly the kind he described.

2. Courts adjudicating limits: Starting on January 20, 2025, dozens of lawsuits challenged DOGE and related executive orders for impoundment of appropriated funds, dismantling USAID, and large‑scale reductions in force. (en.wikipedia.org) On May 9 and May 30, 2025, the AFGE v. Trump case produced a TRO and then a Ninth Circuit ruling that the president lacks authority to direct broad government‑wide RIFs and reorganization without congressional authorization, explicitly drawing a line between what can be done by executive order and what requires Congress. (clearinghouse.net)

3. Turn to Congress: As these legal concerns mounted, reporting describes the White House preparing formal rescission requests under the Impoundment Control Act and sending Congress a package to cancel roughly $9.3 billion in funding for PBS, NPR and State Department accounts, and then advancing the One Big Beautiful Bill Act with over a trillion dollars in targeted spending cuts through the House beginning in late April and May 2025. (cbpp.org)

Taken together, within a few months of the January 31 episode there was an aggressive cutting drive, major court rulings on its legality, and an ensuing push for legislative cuts, matching his prediction in substance and timing.

Using the Federal Reserve’s official DGS30 series (30‑year constant‑maturity Treasury, daily), the yield on January 31, 2025 was 4.76%. Over the subsequent months it rose but peaked at 5.08% on May 22, 2025, never reaching the ~5.5% level Friedberg relayed. After that peak, the yield trended lower, standing around 4.64% by November 26, 2025, well below both the 5%+ highs and the 5.5% target. (fred.stlouisfed.org)

Contemporaneous market coverage describes this same move as the 30‑year yield surging a bit above 5%—around 5.08–5.09%—its highest close since 2023, again with no mention of it approaching 5.5%. (barrons.com) Because the cycle has clearly seen a rise and then a material decline without ever touching ~5.5%, Friedberg’s quoted prediction that the 30‑year yield would go to about 5.5% “before it goes down” did not come true.

As of November 30, 2025 (about 10 months after the January 25, 2025 podcast), the direction of events is consistent with Friedberg’s claim—there is a large-scale build‑out of AI data centers and advanced chip fabs in the U.S.—but his prediction explicitly concerned what would happen “over the next several years,” which has not yet elapsed.

On the facts: the U.S. is seeing huge chip and data‑center expansion. TSMC has announced that total U.S. investments will reach about $165 billion, including multiple additional fabs, advanced packaging plants, and an R&D center in Arizona, with a long‑term build‑out schedule through at least 2030. (azcommerce.com) Arizona and surrounding states are being positioned as a major semiconductor hub, with dozens of semiconductor-related projects and TSMC’s Arizona fabs already producing advanced chips (including leading AI processors) while more capacity is planned. (economy.ac) In parallel, hyperscale cloud providers are sharply increasing capex for AI servers and data centers: TrendForce estimates combined capex of the eight largest CSPs will jump to about $420 billion in 2025 and exceed $520 billion in 2026, driven heavily by GPU-based AI infrastructure build‑outs. (finance.yahoo.com) Major individual projects like Amazon’s recently announced $50 billion in new U.S. data centers for AI and high‑performance computing for federal customers reinforce the scale of ongoing expansion. (wsj.com)

However, Friedberg’s testable claim was that this “data center and chip infrastructure effort” would “obviously” go forward with or without government involvement over the next several years. To judge that, we would need to see whether the build‑out continues (or would have continued) regardless of evolving U.S. federal policy through a multi‑year window—something we cannot yet fully observe only 10 months later. In addition, many of the marquee chip investments explicitly intertwine with U.S. federal policy and incentives (e.g., CHIPS Act–related support for TSMC’s Arizona expansion), which further complicates any present attempt to separate what’s happening because of government from what would happen without it. (techpowerup.com)

Because the timeframe of “several years” has not passed and the counterfactual of “without government involvement” cannot yet be meaningfully evaluated, the prediction is best classified as inconclusive (too early to tell) rather than clearly right or wrong.

The prediction’s key outcome is that “over the coming decade” (i.e., roughly 2025–2035) the U.S. will be severely disadvantaged relative to China in electricity costs and in its ability to deploy AI competitively, if it fails within a few months of January 2025 to stand up an industrial and regulatory engine for advanced nuclear (e.g., pebble‑bed) deployment.

As of November 30, 2025, fewer than two years of that ~10‑year window have elapsed. Even if we try to assess the antecedent (whether the U.S. has meaningfully “stood up” such an engine), the prediction’s actual claim is about the resulting disadvantage over the decade, which cannot yet be measured: relative electricity costs and AI competitiveness vs. China will depend on policy, build‑out, technology and market evolution through the early‑to‑mid 2030s.

Because the timeframe of the forecast extends well beyond the current date, there is not enough observable data yet to determine whether the U.S. will in fact end up “hugely disadvantaged” in the specified ways. Therefore the correctness of the prediction cannot currently be evaluated.

The prediction is about future reconstruction timelines: that many homes destroyed in the January 2025 Los Angeles wildfires will not be rebuilt for 6–7 years, i.e., until roughly 2031–2032. From a January 2025 starting point, 6–7 years later is 2031–2032, which is still in the future relative to the current date of November 30, 2025.

Only about 10 months have elapsed since the fires, so there is no way yet to empirically verify whether a substantial fraction of affected homes will indeed take 6–7 years to be fully rebuilt. Even if we can find information on early rebuilding progress or projected delays, those are forecasts, not actual outcomes observed in 2031–2032.

Because the prediction is explicitly about outcomes years past our current date, it is too early to determine whether it is right or wrong. Therefore the correct status classification is "inconclusive" (too early).

Public reporting shows that the federal TikTok sell‑or‑ban law (the Protecting Americans from Foreign Adversary Controlled Applications Act) did take effect on January 19, 2025, after the Supreme Court upheld it on January 17, 2025, rather than being delayed or mooted by a deal. (en.wikipedia.org) TikTok’s Supreme Court challenge in TikTok, Inc. v. Garland failed, and the company itself stated it would have to shut down without non‑enforcement assurances.

In response, TikTok shut down its U.S. service on January 18, 2025, hours before the deadline, and Apple and Google removed TikTok and other ByteDance apps from their U.S. app stores in anticipation of the law taking effect. (theguardian.com) According to the post‑case summary, TikTok service began to be restored on January 19 after President‑elect Trump signaled he would extend the deadline, but as of January 24, 2025, neither Apple nor Google had restored TikTok to their app stores. (en.wikipedia.org)

Friedberg’s normalized prediction had two key parts:

1. A deal or formal delay around the January 19 deadline so TikTok would not actually be banned.
2. TikTok remaining active in U.S. app stores when the episode aired (within a few days of Jan 18, 2025).

In reality, there was no timely formal deal or legal delay before the January 19 deadline; the ban provisions were allowed to take effect. TikTok was removed from U.S. app stores on or before the deadline and was still absent several days later, even after partial service restoration. (en.wikipedia.org) Therefore, both the “no ban” and “still in app stores” conditions of the prediction were not met, so the prediction is wrong.

Trump’s second term began on Jan 20, 2025, so the first six months ran roughly through July 20, 2025. (en.wikipedia.org) During that period, US–China relations were dominated by an escalating trade war and the “Liberation Day” tariff package, which raised U.S. tariffs on Chinese imports up to 145%, prompting Chinese retaliation and rare‑earth export restrictions. (en.wikipedia.org)

On May 12, 2025, the two sides reached an initial Geneva accord that temporarily reduced tariffs (from 145% to 30% on Chinese goods and from 125% to 10% on U.S. goods) and created a 90‑day truce to allow further negotiations. This was explicitly framed by officials and media as an initial or temporary deal, not a comprehensive settlement; steep tariffs, export controls, and broader strategic disputes remained. (whitehouse.gov) Analysts at the time specifically argued that a broad US–China “grand bargain” was not on the horizon and that only more modest, issue‑specific arrangements were realistic. (scmp.com)

Through late July 2025, coverage continued to describe the situation as a limited truce with ongoing talks and anticipated only narrow deals (for example around fentanyl), not a sweeping grand deal that reset the overall relationship. (en.wikipedia.org) A more substantial multi‑year U.S.–China trade and economic agreement was only announced in early November 2025—over nine months into Trump’s term, outside the six‑month window specified in the prediction. (ustr.gov)

Because no widely recognized, broad “grand deal” that clearly eased overall tensions and created a durable, mutual economic reset was reached within the first six months of the Trump administration, the prediction did not come true as stated.

Donald Trump was inaugurated for his second term on January 20, 2025, so the first six months of his administration run roughly through July 20, 2025.(en.wikipedia.org)

Within that window, on May 12, 2025, the U.S. and China announced a Geneva trade deal and joint statement that dramatically rolled back the new, extremely high reciprocal tariffs that had been imposed in April. The agreement lowered the reciprocal tariff component on each side from around 125–145% to 10% for an initial 90-day period, cutting the U.S. general tariff rate on Chinese goods associated with that measure from 125% down to 10% (though a separate 20% “fentanyl” tariff and pre‑existing Section 301/232 tariffs remained).(whitehouse.gov) This is a very large reduction in the marginal tariff burden that had just been imposed on Chinese imports, i.e., a clear move to “minimize tariff impacts” relative to the peak levels Trump had created.

On the Chinese side, the same Geneva deal required Beijing to cut its tariffs on U.S. goods to 10% and to suspend or remove non‑tariff countermeasures and retaliatory duties that had been put in place after April 2, 2025.(whitehouse.gov) Rolling back those retaliatory tariffs and non‑tariff restrictions materially improves access for U.S. exporters and firms in the Chinese market compared with the pre‑deal situation.

Later, on November 1, 2025, a follow‑on Trump–Xi agreement went further: China committed to “open China’s market to U.S. soybeans and other agricultural exports,” end retaliation against U.S. semiconductor manufacturers and other companies, suspend wide‑ranging retaliatory tariffs and non‑tariff measures, and extend tariff‑exclusion processes, while the U.S. further reduced certain China‑related tariffs and extended its own exclusions.(whitehouse.gov) This later deal reinforces the same basic trade‑off: U.S. tariff relief in exchange for greater Chinese-market access for U.S. entities.

There is nuance—overall U.S. duties on Chinese goods remain elevated because Trump also imposed a general 10% baseline tariff and separate fentanyl‑related tariffs, and many Section 301/232 measures stayed in force.(businessinsider.com) But the prediction was about the structure of an early Trump‑term U.S.–China “grand deal”: the U.S. easing tariff pressure on Chinese imports in return for better access to China for U.S. firms. The May 12 Geneva agreement, occurring well within the first six months, matches that pattern closely—large negotiated cuts to new reciprocal tariffs paired with China rolling back retaliatory tariffs and other barriers on U.S. goods and companies.

Given that this core dynamic did in fact occur, the prediction is best classified as right, albeit with the caveat that much of the relief was partial and initially time‑limited rather than a complete, permanent dismantling of Trump’s China tariffs.

The prediction is framed as a multi‑year structural claim (“over the coming years” the ultimate burden will fall predominantly on taxpayers), but less than a year has passed since the Jan 11, 2025 episode, so it’s too early to evaluate how the cost-sharing equilibrium will ultimately settle.

Current evidence is mixed:

• Historical analyses show that in some years taxpayers already shouldered a very large share of climate‑disaster costs. For example, for 2012, NRDC estimated that private insurers covered only about 25% of climate‑related disaster damages, with the U.S. government paying more than three times as much as private insurers, implying a heavy taxpayer burden in that year.​ (nrdc.org)
• But more recent, disaggregated work from the Federal Reserve on 2021 U.S. disaster damages (~$153B) finds the largest share was borne by property insurers (48%), followed by uninsured/underinsured owners and businesses (37%), while federal government (11%) plus state/local (4%) together accounted for about 15%—well short of a taxpayer‑dominant share.​ (federalreserve.gov)
• For 1H 2025, Aon data summarized by PreventionWeb indicate that a majority of weather‑related losses were absorbed by insurers, with the global insurance protection gap at a record low ~38%, meaning insured losses exceeded uninsured ones; the U.S. drove much of this result. This again suggests insurers, not taxpayers, were carrying most direct financial losses in that period.​ (preventionweb.net)
• A Bloomberg Intelligence analysis (reported mid‑2025) notes that federal spending previously covered up to about one‑third of climate‑related costs but has recently fallen to around 2%, implying the federal taxpayer share, at least in that metric, has declined rather than moved toward dominance.​ (insurancejournal.com)

These data show that (a) taxpayers already sometimes bear very large costs, and (b) in the last few years, insurers and uninsured owners have generally borne larger direct loss shares than government. But because the prediction is explicitly about how the balance will evolve over the coming years, and cost-sharing is still in flux, there is not yet enough elapsed time or clear directional evidence to say that taxpayers will ultimately bear the predominant share. As of Nov 30, 2025, the claim cannot be decisively judged, so the outcome is inconclusive (too early).

Key pieces of Friedberg’s scenario have partially materialized, but the full chain he described has not clearly played out by late 2025.

1. Big losses and stress on state‑backed insurers
– California’s FAIR Plan (insurer of last resort) took massive losses from the January 2025 Los Angeles wildfires and has paid out billions in claims, ending up with an estimated ~$800m deficit and requiring a $1b assessment on member insurers. (reuters.com)
– Florida’s Citizens Property Insurance Corporation projects a record net underwriting loss for 2024 after Hurricanes Helene and Milton, but still expects to rely on its surplus, reinsurance, and (if needed) policyholder surcharges and emergency assessments, not an immediate state budget bailout. (spglobal.com)

2. Who is actually paying the shortfalls?
– In California, the FAIR Plan is explicitly not funded by taxpayers; it is a pool of private insurers. Deficits are handled via assessments on those insurers and temporary surcharges on policyholders statewide (roughly a one‑time ~$60 charge per homeowner for the current assessment), rather than through general tax revenues. (gov.ca.gov)
– California did pass a $2.5b state disaster‑relief package after the LA fires, but this spending is for emergency response and rebuilding support, not a direct bailout of underpriced state insurance schemes. (apnews.com)
– Florida law similarly contemplates Citizens covering post‑storm deficits first via surcharges on its own policyholders and, if needed, emergency assessments on most property‑and‑casualty policies statewide. Those are effectively levies on insureds, not general‑fund taxpayer bailouts, and there is no report yet of a 2024–25 Citizens deficit being covered by state tax dollars. (citizensfla.com)

3. Taxpayer backlash & legislative reforms aimed at ending subsidies
– There is visible consumer anger in California about FAIR Plan surcharges that would be borne by homeowners far from the fire zones, and lawmakers have scrambled for alternatives, but this is framed as opposition to statewide premium surcharges and insurer bailouts—not a broad taxpayer revolt over general‑fund spending. (latimes.com)
– California has moved ahead with substantial insurance reforms: the Sustainable Insurance Strategy and new catastrophe‑modeling rules, FAIR Plan modernization and financing tools, and a 2025 law to create a transparent public wildfire catastrophe model. These changes do tend to allow more risk‑based pricing and seek to stabilize the market, but they were largely initiated in 2023–24 and are not clearly presented as a direct reaction to taxpayers objecting to subsidizing others’ home values. (insurance.ca.gov)
– Florida’s big structural reforms—shifting litigation rules, lifting Citizens’ rate cap on non‑primary residences, and aggressively depopulating Citizens—were enacted in 2022–23, before Friedberg’s January 2025 prediction. In 2025 the state actually expanded homeowner subsidies through new funding for the My Safe Florida Home grant program, which helps pay for wind‑mitigation upgrades. (flgov.com)

4. Time horizon
Friedberg said that “at the end of the day” governments would have to “step in and pay the difference,” provoking taxpayer backlash that would drive reforms to reduce or end subsidies and push prices toward full risk‑reflective levels. Less than a year later, we do see:

• extremely costly disasters in California and Florida;
• severe financial pressure on state‑backed insurers; and
• ongoing regulatory and legislative changes that move toward more risk‑based pricing.

But as of late November 2025, there has not yet been a clear, large‑scale taxpayer‑funded bailout of these insurance schemes, nor an identifiable, taxpayer‑driven political moment that directly produced major subsidy‑cutting reforms. Because the underlying dynamics are long‑running and could still produce the scenario he described over the next several years, it is too early to definitively label the prediction as right or wrong.

As of November 30, 2025, the timeframe for the prediction (“by the end of 2025”) has not fully elapsed, and the 2026 U.S. midterm fields—especially for the House—are still forming. Many candidates have not yet filed, and there is no comprehensive, comparable age dataset across cycles, so we cannot reliably measure whether the 2026 candidate pool is “noticeably younger” than prior midterms.

There are visible signs of a generational-change push and some prominent younger entrants. For example, 36‑year‑old Texas state Rep. James Talarico launched a 2026 U.S. Senate campaign explicitly framed around “generational change” and presenting a youthful contrast to older Republican leadership, getting national write‑ups and viral attention.(statesman.com) Within the Democratic Party more broadly, major coverage has emphasized a post‑2024 push for younger, more relatable leaders, and the retirements of senior figures like Sen. Dick Durbin, Tina Smith, Jeanne Shaheen and others have been framed as opening the door to a younger Senate caucus.(time.com) High‑profile primary challenges such as Rep. Seth Moulton (47) running against 79‑year‑old Sen. Ed Markey in Massachusetts also center explicitly on “generational change.”(washingtonpost.com)

At the same time, many of the most important 2026 Senate prospects for Democrats—like Sherrod Brown (72), Janet Mills (79), and Roy Cooper (68)—are themselves well over 65, and reporting notes that Democrats’ hopes of regaining the Senate majority still depend heavily on these older candidates.(en.wikipedia.org) Across early candidate lists for 2026 Senate and House races, there is a mix of long‑tenured incumbents in their 60s–80s and newer figures in their 30s–40s, but no clear, quantifiable evidence yet that the overall field is younger than in prior midterm cycles.

Given (1) the remaining time in 2025, (2) incomplete candidate slates, and (3) the lack of robust cross‑cycle age data, it’s too early to definitively say whether Friedberg’s prediction about a noticeably younger emerging 2026 field has come true, even though there are qualitative signs pointing in that direction. Therefore the prediction is best classified as inconclusive at this point.

Across 2025, the balance of power in U.S. Republican and executive‑branch foreign policy clearly shifted toward the JD Vance / “America First” restraint camp and away from the traditional pro‑Ukraine, pro‑intervention neoconservatives, with multiple concrete setbacks for the neocon position.

Concrete policy/legislative setbacks for the neocon line

• JD Vance became vice president and a central architect of foreign policy, explicitly skeptical of continued Ukraine aid and favoring economic deals over long‑term security guarantees, which is at odds with neocon priorities of robust, alliance‑based military support. (reuters.com)
• The Trump–Vance administration first paused and then effectively ended U.S. funding for Ukraine’s war effort, pressuring Kyiv toward a negotiated settlement; this is a direct defeat for the neocon goal of sustained, large‑scale U.S. military assistance to Ukraine. (cnbc.com)
• Trump’s Executive Order 14169 imposed a sweeping 90‑day pause on most foreign development aid, and the Rescissions Act of 2025 cut nearly $8 billion from international assistance programs. Traditional neoconservatives have generally supported foreign aid as a tool of American influence, so these moves represent material reversals of that agenda. (en.wikipedia.org)
• On Ukraine diplomacy, an Army secretary seen as a proxy for JD Vance has led a controversial peace initiative closely aligned with Vance’s more accommodationist vision, sidelining more hawkish voices; reporting emphasizes this as evidence of Vance’s “significant influence” over policy. (theguardian.com)
• Prominent neocon figures like Liz Cheney and Bill Kristol publicly describe themselves as out of power and explicitly frame JD Vance’s rise as a repudiation of their interventionist approach inside the GOP, underscoring their loss of intra‑party influence. (breitbart.com)

Countervailing evidence and why the prediction still holds The Trump–Vance administration has pursued very hawkish actions in some theaters—major arms sales and maximalist proposals on Gaza, airstrikes in Yemen, and expanded strikes against alleged traffickers around Venezuela—showing that “America First” is not pure pacifism. (en.wikipedia.org) But these operations are framed in transactional or narrowly defined security terms, not in the neocon vocabulary of democracy promotion and open‑ended commitments, and they coexist with a decisive rollback of the flagship neocon project of this era: arming and financing Ukraine at scale.

Given (1) the elevation and clear policy impact of JD Vance and like‑minded figures, (2) the termination of Ukraine war funding and cuts/pauses to foreign aid, and (3) the explicit marginalization acknowledged by leading neocon voices, the specific prediction—that the pro‑war neocon faction would lose influence relative to a Vance‑aligned restraint camp, with concrete setbacks for their agenda in 2025—has been borne out.

Evidence from 2025 shows a clear surge in humanoid-robot R&D, investment, and pilot deployments, but not the level of mainstream commercial use and broad public recognition as a major platform that the prediction requires.

Surge and visibility (prediction gets this part roughly right):

• Multiple outlets explicitly described 2025 as a pivotal year for humanoid robots. Wired called it the year that multipurpose humanoid robots "go commercial," highlighting Boston Dynamics’ Atlas entering a Hyundai factory and Agility’s Digit and Figure’s robots beginning paid work in warehouses and other sites. (wired.com)
• Business Insider’s late‑2025 overview of six leading humanoid-robot companies notes that Agility’s Digit is already deployed commercially at GXO Logistics and Amazon, and frames humanoids as rapidly moving from science fiction toward practical reality. (businessinsider.com)
• Governments and industry coalitions treated humanoids as strategic: China’s government work report and trade-fair materials describe 2025 as the first year of mass production for humanoid robots and integrate "embodied intelligence" into national industrial policy; South Korea launched the K‑Humanoid Alliance to build a national humanoid ecosystem. (globalomp.com)
• China hosted the first World Humanoid Robot Games in August 2025 with 500+ robots competing, and conferences like the Humanoids Summit expanded globally, underscoring growing public and investor visibility. (en.wikipedia.org)

But deployment is not yet mainstream commercial use (prediction fails here):

• The scale of real-world deployments remains small and experimental. GXO’s own disclosures describe its Digit deployment at a Spanx warehouse as the first commercial application of a humanoid robot and emphasize that only a small fleet is in use; Business Insider notes GXO has just two Digit units active and that widespread deployment "is not imminent." (therobotreport.com)
• A 2025 industry white paper on humanoid applications characterizes 2025 as a “critical starting phase” for industrial deployment, but explicitly projects 2027 as the year of small‑batch production and 2030 as the starting point for real commercialization, implying that 2025 is still pre‑mainstream. (sohu.com)
• Chosun Ilbo’s 2025 analysis likewise calls this the inaugural year of mass production but quotes Unitree’s CEO predicting a "ChatGPT moment" for humanoids 2–3 years in the future, and cites Goldman Sachs’ forecast of a relatively small $1.5B market in 2025 versus $37.8B by 2035—again suggesting that broad, mainstream commercial penetration has not yet arrived. (chosun.com)
• Even where commercial pilots exist (e.g., Digit at Amazon and GXO, Apptronik’s Apollo at Mercedes factories, Kepler’s K2 robots entering mass production), the coverage consistently describes them as early trials or milestones rather than ubiquitous, routine tools across warehouses, factories, or consumer settings. (reuters.com)

Public recognition as a “major technology platform” is still emerging, not achieved:

• Media, investors, and policymakers are heavily focused on humanoids, and 2025 saw new summits, national alliances, and even international robot games, which indicates rising awareness. (en.wikipedia.org)
• Yet the same industry reports and news pieces frame humanoids as on the verge of breakthrough rather than already widely adopted—explicitly forecasting a ChatGPT‑like popularization in several years and placing true large‑scale commercialization closer to the late 2020s. (chosun.com) There is no evidence that the general public in 2025 sees humanoid robots as a mature, mainstream platform on par with smartphones or even consumer AI assistants.

Because by late 2025 humanoid robots are high-visibility but still in early, small‑scale commercial pilots, and industry consensus pushes genuine mainstream deployment and broad public recognition into the later 2020s, Friedberg’s stronger claim—that by year‑end 2025 humanoid robots would be in mainstream commercial use cases and widely recognized by the public as a major new technology platform—has not come true.

Evidence on this multi‑part prediction is mixed:

• Legacy primes’ stock performance in 2025:

  • Boeing (BA): After a steep 32% drop in 2024 and an additional 9% slide early in 2025, Boeing’s shares recovered somewhat; as of late November, they are only up “mid‑single digits” year‑to‑date and remain volatile, with ongoing regulatory and production issues. This lags the S&P 500’s ~16% gain in 2025. (barrons.com)
  • Lockheed Martin (LMT): By early July, LMT was down about 4% YTD and by mid‑October it had a one‑year return of around –19%, materially underperforming the S&P 500. Lockheed also cut its 2025 profit outlook due to F‑35 technology‑upgrade delays and losses on fixed‑price classified programs, clear business headwinds. (benzinga.com)
  • RTX (Raytheon): In contrast, RTX has had an excellent 2025. StatMuse and Macrotrends data show its 2025 return in the ~37–50%+ range, far ahead of the market. The company reports a record ~$218 billion backlog and “unprecedented” missile demand, with 4–6% expected sales growth in 2025. (statmuse.com)
    Net: Two of the three cited primes (BA, LMT) did face notable headwinds and at best modest or negative stock performance, but RTX is a major counterexample, enjoying strong fundamentals and outsized returns rather than being “deeply challenged.”

• Performance of newer tech‑driven defense firms:

  • Palantir (PLTR): Multiple sources show PLTR as one of the best‑performing U.S. defense‑adjacent stocks in 2025. A Benzinga comparison table has Palantir up ~79% YTD by early July, versus Boeing ~26%, RTX ~26%, and Lockheed –4%. (benzinga.com) Later analysis notes Palantir up ~75% YTD by early July and roughly 300% YTD by late October, vastly outpacing both the legacy primes and the S&P 500. (nasdaq.com) Over the past year through November, PLTR is up about 150%. (investing.com)
  • Anduril: Anduril raised about $2.5 billion in 2025 at roughly a $30–30.5 billion valuation and is reported to have more than $2 billion in active government contracts. It won high‑profile programs such as Australia’s A$1.7 billion Ghost Shark XL‑AUV fleet, is a contractor on the U.S. Space Force’s Golden Dome missile‑defense effort, and its Fury (YFQ‑44) drone is one of the winning designs in the USAF’s Collaborative Combat Aircraft program. (businessinsider.com) These wins and its funding trajectory support the idea that tech‑centric “new primes” are gaining momentum.

• Shift in defense spending and contracting patterns:

  • Policy and funding signals clearly favor AI‑ and software‑heavy solutions. Commentators note firms like Palantir and Anduril winning multi‑hundred‑million‑dollar data, AI, and autonomy contracts and becoming central to a “new breed” of defense tech suppliers. (aoav.org.uk) Venture funding into defense tech startups roughly doubled in 2025 to over $19 billion, driven by Pentagon modernization priorities around AI, autonomy, and software‑defined systems. (businessinsider.com)
  • However, at Baird’s 2025 Defense & Gov’t conference, Anduril’s own leadership emphasized that the Department of Defense still isn’t procuring at scale from new entrants—most of their work is in prototypes and early fielding, not yet in large production programs with the same volume and stability that legacy primes enjoy. (wash100.com) This suggests the “swing” of spending away from incumbents toward startups is underway but far from complete.

Putting it together:

• The relative stock‑performance call—that newer tech‑driven defense companies like Palantir would dramatically outperform legacy primes—did happen in 2025. Palantir’s returns tower over Boeing, Lockheed, and even the strongly performing RTX. (benzinga.com)
• The claim that legacy primes would be “deeply challenged…across the board” is only partially supported. Boeing and Lockheed faced real operational and earnings headwinds plus subpar share performance, but RTX experienced record demand, strong financial results, and very strong stock gains—directly contradicting a blanket “headwinds” thesis. (marketwatch.com)
• The idea of a wholesale, near‑term shift of government defense spending away from primes toward ROI‑driven startups is also overstated. While policy, venture funding, and some major contracts favor AI/autonomy players, official statements from both the Pentagon and new‑entrant executives indicate that large‑scale production dollars still overwhelmingly flow to the traditional primes, with new firms mostly in prototype and early deployment stages. (defensenews.com)

Because the prediction bundles together (1) a relative stock‑market bet that largely worked, (2) an industry‑wide “deep challenge” claim that is contradicted by RTX’s strong 2025, and (3) a structural spending‑shift claim that is only partially and gradually playing out, the overall truth value depends heavily on which component you weight most. The data are sufficient, but they support some parts and clearly contradict others, so the outcome for the prediction as stated is best characterized as ambiguous rather than cleanly right or wrong.

OpenAI spent 2024–25 working with regulators on a plan to convert its capped‑profit subsidiary into a standard equity structure and for‑profit public benefit corporation (PBC), while facing lawsuits and scrutiny over the treatment of its nonprofit assets. (arstechnica.com)

On 28 October 2025 OpenAI announced and implemented a new structure: the original nonprofit became the OpenAI Foundation and the operating business became OpenAI Group PBC, a for‑profit corporation with ordinary stock, majority investor ownership and a path to an eventual IPO, while the foundation retained a large equity stake and the power to appoint and remove all directors. (openai.com) Regulators in Delaware and California formally approved this restructuring, and major outlets reported that OpenAI had completed its conversion to a for‑profit PBC. (apnews.com)

Although litigation and policy debates continue, no court has invalidated the restructuring, and the new for‑profit PBC structure is in force as of late 2025. (theverge.com) That means OpenAI was in fact able to complete a legally approved transition to a conventional for‑profit corporate vehicle during 2025, so Friedberg’s prediction that it would not be able to convert from nonprofit to for‑profit in 2025 is contradicted by events.

By late 2025, multiple large‑scale, multi‑billion‑dollar packages for U.S. hardware manufacturing had been announced, matching the prediction’s direction. In April 2025, the White House reported over $5 trillion in new U.S.-based investments in Trump’s first 100 days, emphasizing a series of blockbuster advanced‑manufacturing commitments: Apple’s $500 billion U.S. manufacturing and training plan, TSMC’s additional $100 billion for U.S. chip manufacturing, Johnson & Johnson’s $55 billion, Novartis’s $23 billion to build or expand ten U.S. plants, and Hyundai’s $21 billion U.S. investment including a $5.8 billion new steel plant—each framed as major U.S. manufacturing build‑outs rather than purely digital or services spending.(whitehouse.gov)

In parallel, semiconductor‑focused industrial policy continued to drive enormous hardware manufacturing capex. By 2025 the Commerce Department had distributed over $30 billion in CHIPS Act subsidies across 17 firms (including Intel, TSMC, Samsung, Micron), catalyzing more than $100 billion in new U.S. fab projects.(ainvest.com) TSMC specifically expanded its Arizona plans to a $65+ billion complex and, alongside earlier awards, received a $6.6 billion U.S. subsidy plus up to $5 billion in loans, while separately announcing an additional $100 billion U.S. chips‑manufacturing program and a multi‑fab Arizona “gigafab” cluster.(investing.com) These are clearly multi‑billion financing/subsidy‑linked build‑outs of physical manufacturing capacity in the U.S.

On the robotics/“physical AI” side, 2025 also saw U.S.-based manufacturing expansions tied directly to robotics: Figure AI announced BotQ, a high‑volume U.S. facility intended to manufacture 12,000 humanoid robots per year; Nvidia and Foxconn launched a new Houston AI manufacturing campus where humanoid robots will assemble Nvidia AI servers; and SoftBank unveiled “Project Crystal Land,” a proposed trillion‑dollar‑scale AI and robotics industrial hub in Arizona in partnership with TSMC, explicitly framed as a U.S. manufacturing center for AI and robotics hardware.(en.wikipedia.org) While not every robotics project discloses a precise multi‑billion capex figure, they are part of the same wave of highly publicized industrial‑scale hardware and robotics build‑outs.

Taken together, these 2025 announcements constitute several multi‑billion‑dollar financing and subsidy–driven build‑outs of U.S. hardware (and related robotics) manufacturing. Their headline sizes (tens to hundreds of billions, with some projects framed at the trillion‑dollar level) and the way they dominate policy and business news place them in the same general league of scale and prominence as the 2023–24 AI compute megaprojects, so Friedberg’s prediction is best judged as having come true in substance.

Available reporting through late 2025 shows no major new capital or corporate event for Waymo in 2025 of the type Friedberg described (large external financing round, IPO, or merger/acquisition).

• Waymo’s last major capital raise was a $5.6 billion Series C funding round announced on October 25, 2024, led by Alphabet with participation from outside investors such as Andreessen Horowitz, Fidelity, Silver Lake, Tiger Global, and T. Rowe Price. This is consistently dated to October 2024, not 2025. (cnbc.com)
• Follow‑on coverage in 2024–2025 continues to reference that same $5.6 billion Series C as Waymo’s big round, putting its valuation above $45 billion, with no subsequent 2025 financing round disclosed. (gurufocus.com)
• Throughout 2025, news about Waymo focuses on geographic expansion and partnerships (e.g., expansion of robotaxi services in Silicon Valley, Austin, and Miami; partnerships with Uber and Lyft in markets like Austin and Nashville), but these are commercial/operational agreements, not mergers or acquisitions. (en.wikipedia.org)
• As of late November 2025, major outlets still describe Waymo as Alphabet’s self‑driving unit/subsidiary, indicating no IPO or spin‑off and no sale to or merger with another company. (axios.com)

Because the prediction was specifically “this year” (2025) and pointed to a new massive financing, IPO, or M&A event, and none of those occurred in 2025 (the big funding round happened in 2024), the prediction is best scored as wrong.

Available evidence shows that Waymo did not complete a new “major capital markets transaction” in 2025.

Waymo’s last large external raise was a $5.6 billion Series C funding round that closed on October 25, 2024, led by Alphabet with participation from major institutional investors; this round is consistently described in contemporary coverage and transaction records as a 2024 deal, not 2025. (cnbc.com) Overviews of Waymo’s financing history as of March–June 2025 list total funding of about $11.1 billion and identify that 2024 Series C (plus earlier 2020 and 2021 rounds) as the latest funding, with no additional 2025 equity or debt rounds recorded. (en.wikipedia.org) Analysts discussing a possible spinoff or separation from Alphabet in 2025 treat it purely as speculation and explicitly reference the 2024 Series C as the most recent capital raise. (forbes.com)

At the same time, 2025 news about Waymo focuses on operational expansion and commercial partnerships (e.g., launching services via Uber and Lyft in cities like Austin and Nashville), but these are structured as service/operations agreements, not mergers or acquisitions of, or by, major ride‑hailing companies. (en.wikipedia.org) Waymo also remains a wholly owned (and non‑public) subsidiary of Alphabet throughout 2025, with no IPO or spin‑off completed by November 30, 2025. (en.wikipedia.org)

Because the prediction specifically required that in 2025 Waymo would execute a major capital‑markets event (large financing, IPO, or M&A with a big ride‑sharing company), and no such 2025 transaction occurred, the prediction is wrong.

Evidence from 2025 strongly indicates that socialist movements in the U.S. did grow substantially in both membership and political influence relative to 2024.

1. Organizational membership (breadth): The Democratic Socialists of America (DSA), the largest socialist organization in the U.S., saw membership fall to about 64,000 by October 2024, but then experience a sharp rebound to over 80,000 by October 2025 and over 90,000 by November 2025—roughly a 40% increase from the 2024 trough in a single year. (en.wikipedia.org) A September 2025 analysis of the DSA convention likewise reports the group at “more than 80,000 members” and on track to possibly exceed 100,000. (solidarity-us.org)

2. Youth and campus organizing: Young Democratic Socialists of America (YDSA) reported 47% growth in chapter membership since November 2024, indicating rapid expansion among student and youth organizers. (notus.org) An October 2025 Axios/Generation Lab poll found U.S. college students now view socialism more favorably than capitalism (34% vs. 17%), suggesting a widening base of young sympathizers. (axios.com)

3. Electoral and political influence (depth): DSA‑aligned officeholders expanded from “over 200 elected officials” in early 2024 to more than 250 by 2025. (en.wikipedia.org) NYC‑DSA and DSA nationally were central to Zohran Mamdani’s 2025 New York City mayoral win, widely described as the biggest electoral victory in DSA history and a major boost to democratic socialist influence in a flagship U.S. city. (en.wikipedia.org)

Taken together—large year‑over‑year membership gains from a 2024 low point, strong growth among youth and campus groups, and increased electoral clout—the data support Friedberg’s prediction of a “dramatic rise” in socialist movements in the U.S. during 2025 compared with 2024.

As of November 30, 2025, there is no record that Jeff Bezos has sold The Washington Post. Major coverage of the paper throughout 2025 continues to describe it as owned by Jeff Bezos, including Reuters on January 7, 2025, reporting layoffs at “the organization, owned by Amazon founder Jeff Bezos,” and referring to no change in ownership. (reuters.com) The Washington Post’s own February 26, 2025 story about shifting its opinion section explicitly calls Bezos the paper’s owner with no mention of a sale. (washingtonpost.com) Business Insider’s May 2025 coverage of buyouts similarly treats Bezos as the current owner during restructuring. (businessinsider.com)

In addition, reporting on speculation about a possible sale consistently notes that Bezos or his representatives say the Post is not for sale, including a New Yorker article in May 2025 stating he recently told an interested buyer that it is not for sale, and earlier coverage of Commanders-related rumors where a Bezos spokesperson said the Post is not for sale. (newyorker.com) There is no credible report of any completed sale in 2025, which would almost certainly attract extensive media coverage.

However, because the prediction specifies “during calendar year 2025” and the year has not yet ended as of the current date (November 30, 2025), it is still theoretically possible—though seemingly unlikely—that a sale could occur in December 2025. Therefore, with the deadline not yet passed, the prediction cannot be definitively graded as right or wrong and is best classified as inconclusive (too early to tell).

Multiple broad Chinese tech / internet benchmarks had strong double‑digit gains in 2025 and clearly outperformed many other equity sectors.

• The KraneShares CSI China Internet ETF (KWEB), a widely used proxy for Chinese tech/internet stocks, shows a YTD return of 35.2% (NAV) as of Oct 31, 2025.【1search4】 A separate performance commentary notes KWEB had returned about 49–50% year‑to‑date later in 2025, highlighting a very strong run.【3search2】
• Broader commentary on Chinese tech confirms a major rally: Axios reports Chinese tech stocks up over 44% year‑to‑date from their April lows, outpacing leading U.S. tech stocks and the “Magnificent 7.”【3news13】 The Financial Times likewise notes the Hang Seng Tech Index entered a bull market in 2025, rising about 25% from its January low and beating the Nasdaq 100, with large caps like Alibaba up ~43% and Xiaomi ~34%.【3news14】
• In contrast, the S&P 500 total‑return YTD as of Oct 31, 2025 is about 17.4%,【2search2】 and the U.S. Technology Select Sector SPDR (XLK)—the best major U.S. sector—shows a 29.9% YTD return over the same period.【2search1】 A sector breakdown for the S&P 500 shows most other sectors (healthcare, financials, materials, real estate, energy, etc.) with mid‑single to low‑20s YTD returns, or less, all below KWEB’s mid‑30s+ YTD gains.【3search0】

Taken together, these data show that broad Chinese tech equities (as represented by ETFs like KWEB and indices like Hang Seng Tech) appreciated significantly in 2025 and outperformed the S&P 500 and the majority of U.S. equity sectors, including most non‑tech sectors. That matches the normalized prediction that Chinese tech stocks would have “a pretty good run” and beat many other sectors, conditional on reasonably supportive macro conditions. Therefore, the prediction is best judged as right.

Available 2025 data show vertical SaaS has generally been one of the stronger parts of software, not among the worst-performing, and there is no clear evidence that AI‑driven in‑house tools and pricing compression uniquely damaged its valuations or revenues.

Key points:

• Valuations and relative performance. Multiple independent valuation and M&A analyses in 2025 find that vertical SaaS companies still trade at a premium to horizontal SaaS and are viewed as especially attractive:

  • AGC Partners’ early‑2025 report shows vertical SaaS trading at about 7.0× EV/Revenue vs. 4.8× for horizontal SaaS, with similarly strong growth and margins, implying investors pay up for vertical focus rather than shun it. (agcpartners.com)
  • A Baker Tilly / AInvest summary of Q3 2025 data likewise reports vertical SaaS firms at 7.0× EV/revenue vs. 4.8× for horizontal, explicitly saying vertical SaaS companies "continue to outperform horizontal peers." (ainvest.com)
  • A 2025 valuation overview notes that AI‑first and vertical SaaS are the most sought‑after categories, often rewarded with 8–12× revenue multiples, while broader horizontal platforms without clear ROI sit around 3–5×. (ful.io)
  • A broad SaaS metrics benchmark (2,000+ companies) highlights "Vertical SaaS outperforming": vertical players are 1.5–3.3× more likely to be outlier winners and account for 54% of Q3 2025 SaaS M&A, with the conclusion "Vertical‑first wins." (rockingweb.com.au)
  • A July 2025 sector note from JM Financial contrasts vertical vs. horizontal SaaS since the 2021 peak: horizontal SaaS endured sharper multiple compression, while vertical SaaS names like AppFolio, Procore, Veeva, Samsara, and Guidewire generally maintained 5–15× EV/Sales, with "more orderly and muted de‑rating" and structurally stronger pricing power due to sticky, mission‑critical workflows. (fr.scribd.com)

• Stock and growth benchmarks do not show vertical SaaS as a bottom tier.

  • A First Analysis quarterly review does show that in the June 2025 quarter, the vertical SaaS group’s shares rose only about 5% vs. 13% for its broader SaaS universe, lagging categories like cybersecurity and data visibility. That is mild underperformance for that quarter, not "worst‑performing software assets" for the year. (firstanalysis.com)
  • The Software Equity Group’s SEG SaaS Index reports that "Vertically Focused" public SaaS names saw EV/revenue medians fall from 4.5× (3Q24) to 3.2× (3Q25) while the overall index went from 5.9× to 5.3×, indicating some relative multiple compression—but again, several other categories (e.g., analytics, sales & marketing) also sit in the low‑multiple range, so vertical is not uniquely or clearly the "worst" segment. (softwareequity.com)
  • By contrast, other 2025 performance roundups emphasize that industry‑focused, vertical platforms are among the winners: a 1H’25 SaaS stock review notes that industry‑focused platforms like Veeva and Guidewire outperformed horizontal solutions, and calls out "Vertical Specialization Advantage" as a key driver. (linkedin.com)
  • A separate SaaStr analysis, "The Vertical SaaS Gold Rush," shows many vertical or non‑tech B2B names (Samsara, ServiceTitan, Toast, Shopify, etc.) growing 2–3× faster than generic enterprise SaaS; most horizontal incumbents like Salesforce are in single‑digit growth. (saastr.com)
  • Multiple benchmark summaries (e.g., OpenView/WeBuildSaaS, LinkedIn analyses) show median growth of ~45% for vertical SaaS vs. ~28% for horizontal, with top‑quartile verticals at ~100%+ YoY growth—again describing vertical SaaS as "crushing" horizontals, not as a laggard. (webuildsaas.com)

• AI’s effect: more tailwind than systemic damage to vertical SaaS.

  • A 2025 vertical‑SMB SaaS benchmark from Tidemark concludes that "Vertical SaaS has arrived" and that AI is now a revenue driver, with companies that adopt AI in their products and GTM seeing stronger growth, rather than mass churn to in‑house tools. (tidemarkcap.com)
  • Coverage of the "third wave" of vertical SaaS (cloud + fintech + AI) from investors like a16z describes AI as expanding vertical SaaS TAM by automating labor‑intensive workflows and enabling higher net revenue retention—again positioning AI as a core growth catalyst for vertical SaaS firms such as Toast, not as a force that is broadly eroding their revenue base. (yourstory.com)
  • Two detailed essays from Reformation Partners (Jan 2024 and Nov 2025) explicitly rebut Friedberg’s thesis that internal AI‑built tools would make vertical SaaS "doomed." They argue, with examples from construction, hospitality, legal, and other sectors, that most vertical SaaS customers lack the in‑house engineering depth to build and maintain comparable tools; that internal builds carry higher ongoing risk and cost than simply buying the vertical standard; and that AI productivity gains accrue to vertical SaaS vendors themselves, allowing them to ship better products or cut prices while preserving margins. The author concludes, nearly two years into the AI boom, that the original “vertical SaaS is dead” prediction has not been borne out. (medium.com)
  • More general commentary on AI and SaaS economics in 2025 acknowledges some margin and pricing pressure across SaaS as AI features become table stakes, but does not single out vertical SaaS as uniquely impaired; instead, it typically notes that commoditization risk is greatest for generic horizontal tools, while domain‑specific, data‑rich platforms retain differentiation. (implicator.ai)

Putting this together:

• Vertical SaaS did not become "among the worst‑performing software assets" in 2025. Across valuations, M&A share, and growth metrics, it is more often described as outperforming horizontal SaaS and as a favored category for investors.
• While there is some evidence of moderate multiple compression and a few quarters of relative underperformance for certain vertical cohorts, that is part of a broader SaaS repricing and does not match the severity or uniqueness implied by the prediction.
• The specific causal story—that enterprises would rapidly adopt AI‑driven in‑house tools, causing widespread churn and pricing compression that disproportionately harms vertical SaaS—has not materialized at scale. Available data and investor commentary instead show AI being integrated into vertical SaaS products and business models, largely as a competitive advantage rather than an existential threat.

On balance, Friedberg’s prediction that vertical SaaS would be one of the worst‑performing software assets in 2025 due to AI‑driven in‑house tools and pricing compression is not supported by the observed 2025 data, so it is best classified as wrong.

Evidence from 2025 shows both regulatory reforms and major new nuclear project announcements in the U.S. that match the prediction:

• Regulatory reforms easing nuclear build‑out: In May 2025, the White House issued an executive order "Ordering the Reform of the Nuclear Regulatory Commission" directing the NRC to facilitate increased deployment of new reactor technologies, lower regulatory and cost barriers, reorganize to process licenses more quickly, and explicitly add facilitation of nuclear power to its mission, not just safety. (whitehouse.gov) In June 2025, the NRC implemented a new fee rule cutting hourly charges for advanced reactor applicants and pre‑applicants by more than 50%, as required by the ADVANCE Act, directly lowering the cost of licensing new reactors. (ans.org) Together, these are material regulatory changes intended to speed and cheapen advanced nuclear licensing.

• Significant new U.S. nuclear power plant projects announced in 2025: In July 2025, Westinghouse publicly laid out plans to build ten new AP1000 reactors in the USA by 2030, an explicit large‑scale build‑out, framed in the context of Trump’s May 23 executive orders to support the U.S. nuclear supply chain and quadruple nuclear capacity by 2050. (world-nuclear-news.org) In October 2025, Amazon and Energy Northwest announced the Cascade Advanced Energy Facility in Washington state, a new SMR power station using X‑energy’s Xe‑100 reactors (initially four units, expandable to 12) to supply Amazon’s data‑center demand—one of the first large, tech‑backed modular nuclear projects in the U.S. (tomshardware.com) In parallel, DOE approved a $1 billion federal loan in 2025 to restart Three Mile Island Unit 1 (the Crane Clean Energy Center) for long‑term power to Microsoft data centers, signaling federal support for bringing dormant nuclear capacity back online as part of a broader build‑out. (apnews.com)

Because 2025 saw both (1) meaningful regulatory reforms at the NRC that lower costs and aim to speed approvals, and (2) high‑profile announcements of new reactors and restarts in the U.S., the prediction that 2025 would bring announcements of nuclear build‑out and/or regulatory easing for nuclear power is substantially borne out.

The prediction concerns conditions by the year 2030 (specifically, about 100,000 operational eVTOL “flying cars” in China used as passenger air taxis). The current date is November 30, 2025, so 2030 has not yet occurred. There is no way to verify now whether this 2030 outcome will be met. Therefore, the correctness of the prediction is too early to determine.

The prediction’s timeframe has not elapsed yet. Friedberg was effectively saying that by roughly seven years after December 20, 2024 (around 2031–2032) it will be standard practice for organizations to use AI systems that can automatically design and generate production software (including security/compliance logic) from natural‑language instructions and analysis of existing software.

As of November 30, 2025:

• Advanced AI coding agents such as Devin can already plan, write, test, and refactor code autonomously, integrate with tools like Slack/Jira/Linear, learn from existing codebases, and generate production pull requests for real companies, including large refactors at a bank‑like fintech (Nubank).(devin.ai)
• These tools are still in early‑access / pilot‑style deployments and are framed as powerful assistants or teammates whose work must be reviewed before deployment, not as universally adopted, fully trusted, compliance‑aware generators of end‑to‑end production systems across organizations.(newbits.ai)

However, the claim is about what will be standard practice by around 2031–2032, and we are only in 2025. Even though we can say the prediction is not yet fully realized, we cannot judge its final correctness because the deadline is several years in the future.

Therefore, the appropriate status is: it’s too early to tell whether the prediction will ultimately be right or wrong.

As of November 30, 2025, Google has not launched a clearly defined ad-supported version of the Gemini AI assistant with a corresponding “pay to remove ads” option.

Evidence:

• On Alphabet’s Q4 2024 and early‑2025 earnings calls, Sundar Pichai repeatedly stated that Gemini does not currently have ads and that monetisation is focused on a free tier plus paid subscriptions, while noting they have “very good ideas” for future native ad concepts in Gemini and will likely explore ads later, not in 2025. (theverge.com)
• Multiple reports summarising those calls reiterate that Gemini is presently monetised through free access and paid tiers, not an ad‑supported vs ad‑free split. (news.aibase.com)
• A November 3, 2025 report says Google is experimenting with ads in AI experiences such as “AI Mode” and the Gemini assistant, and that ads are expected to appear in the free versions of Gemini, but also states that a full roll‑out will not happen in 2025—indicating plans and early tests, not a launched, stable business model. (voice.lapaas.com)
• Current public descriptions of Gemini pricing (as of mid‑2025) list a free tier, a $20/month Gemini/AI Pro tier, and a higher‑priced AI Ultra tier, with no mention of an explicit ad‑supported tier where users pay specifically to remove ads from Gemini chats. (tech.yahoo.com)
• Google has introduced ads into AI-powered search overviews / “AI Mode” in Search (which use Gemini technology), but these are framed as search ads inside AI summaries, not as part of a Gemini assistant freemium ad/no‑ad subscription structure. (impressiondigital.com)

Friedberg’s prediction is about what will eventually happen, with no specific time horizon. The available evidence shows Google moving in the predicted direction (planning and testing native ads, hinting at user “options” similar to YouTube), but not yet delivering a full ad-supported Gemini assistant with a paid no‑ads option. Since the relevant future event could still occur after 2025, it’s too early to label the prediction as either right or wrong.

As of November 30, 2025, no company (including Google) has demonstrated a large‑scale, fault‑tolerant quantum computer capable of practically running Shor’s algorithm at scales that would actually break standard public‑key cryptosystems (e.g., 2048‑bit RSA, standard ECC curves) used on the internet.

Relevant current facts:

• Google’s major public milestones are still in the noisy intermediate‑scale quantum (NISQ) regime. After its 2019 “quantum supremacy” result with the 53‑qubit Sycamore processor, later work has focused on better qubits and error‑correction, not on a full‑scale, cryptography‑breaking device.
• Research papers and Google blog posts since then describe progress on error-corrected logical qubits and improved fidelities, but they do not claim any machine that can run Shor’s algorithm at cryptographically relevant sizes.
• Government and standards bodies are still operating under the assumption that cryptography‑breaking quantum computers do not yet exist, and are preparing via post‑quantum cryptography precisely because such machines are expected only in the future, on an uncertain timeline.

The prediction, however, explicitly allowed 3–7 years from around 2024 (i.e., roughly 2027–2031) for such a machine to exist in principle if Google continues on its trajectory. As of late 2025 we are only about 1–2 years into that window. There is not yet evidence that Google has achieved this, but there is also not yet enough time elapsed to say definitively that it won’t by 2027–2031.

Because the prediction is about capabilities by a future date that has not arrived yet, and current evidence doesn’t falsify the possibility that Google might get there within that 3–7 year window, the correct status today is “too early to tell.”

The prediction has a horizon of ~10 years from the recording date of 7 Dec 2024, i.e., until about 2034. As of today (30 Nov 2025), fewer than 2 years have elapsed, so it is too early to say whether the forecast will prove right or wrong.

Current data show that China already has larger total power generation capacity than the U.S., but not at 4x yet. Various international energy statistics (e.g., from the IEA, U.S. EIA, and other energy data compilers) indicate that in recent years China’s installed power capacity and electricity generation have been on the order of about 2× that of the U.S., depending on the year and metric (installed capacity vs. actual generation), but not 4×. However, this only describes the situation now, not in 2034.

Because the forecast explicitly concerns the situation around 2034, and that date has not yet occurred, the truth value of the prediction cannot be determined at this time. Therefore the status is inconclusive (too early).

Grok 3 was in fact trained on xAI’s Colossus supercomputer at the ~100,000–200,000‑GPU scale (far beyond the ~20–32k‑GPU clusters that previous frontier models used), providing precisely the kind of new high‑compute data point Gavin Baker and Friedberg were talking about. (rdworldonline.com) Multiple analysts explicitly framed Grok 3 as an experiment on scaling laws: Tencent/WallstreetCN’s post‑mortem describes Grok 3’s 100k+ H100 run as showing that pre‑training scaling laws “have not hit a ceiling” and still improve performance, albeit with poor cost‑effectiveness, concluding that “Scaling Law did not hit a wall.” (longbridge.com) Technical explainers and blogs similarly note that Grok 3’s performance gains over Grok 2, given roughly 10x more compute, are largely what standard scaling‑law curves predict—clear improvement but not a dramatic new “generation” jump—indicating continued validity of scaling rather than a breakdown. (uxtigers.com) Ethan Mollick characterized Grok 3 as the first release we know used an order of magnitude more training compute than GPT‑4‑class models and said it would help test whether the first (pre‑training) scaling law still holds; after the results, he wrote that Grok 3 landed right at expectations, with no need to revise the consensus that more compute still buys more capability. (ethanbholland.com) While some online discussions still debate whether we are near a “wall,” the empirical outcome from Grok 3 is broadly read as evidence that scaling laws remain intact at this larger scale rather than having already failed, which is exactly the kind of resolution Friedberg predicted Grok 3 would provide.

The prediction’s time horizon is ~10 years from the podcast release date of 7 Dec 2024. Adding 10 years gives 2034 as the endpoint, which is well beyond the current date of 30 Nov 2025. Since fewer than two years have elapsed, we cannot yet assess whether there will indeed be “another decade of innovation ahead of us,” regardless of whether training scaling laws break. Therefore, the status of this prediction is inconclusive (too early to tell).

By late February 2025, there was substantial but conflicting evidence about whether Grok 3 had truly become the state‑of‑the‑art general‑purpose LLM relative to OpenAI/Microsoft’s best publicly available frontier models.

Evidence that supports the prediction

• xAI launched Grok 3 around February 17–20, 2025, describing it as their new flagship model trained with roughly 10× the compute of Grok 2 and claiming it surpasses OpenAI on benchmarks like AIME (math) and GPQA (PhD‑level science). (es.wikipedia.org)
• xAI’s published benchmarks and coverage in outlets like Beebom and ZeroHedge report Grok 3 outperforming GPT‑4o, Claude 3.5 Sonnet, Gemini 2.0 Pro, and DeepSeek V3 on AIME 2024, GPQA Science, and LiveCodeBench, and also beating OpenAI’s o3‑mini on some reasoning benchmarks. (beebom.com)
• Grok 3 (under the alias “chocolate”) reached the #1 position on the LMSYS Chatbot Arena with an Elo score around 1400–1402, ahead of GPT‑4o and DeepSeek R1, which xAI and supporters framed as evidence it was the top chatbot overall. (twitter.com)

Evidence that cuts against the prediction

• Independent commentary after launch emphasized that Grok 3 was competitive but not clearly dominant. Ethan Mollick (Wharton) described Grok 3 as a “very solid frontier model” but not a clear leader and “not one you would stop using your current frontier model for,” adding that while it beats some OpenAI models on selected benchmarks, it does not clearly surpass OpenAI’s o3. (aol.com)
• Gary Marcus similarly argued that Elon Musk’s promise that Grok 3 would be “the smartest AI ever” was not borne out, calling the launch “no game changer” relative to OpenAI’s best models. (aol.com)
• Comparative write‑ups on Grok 3 vs. OpenAI’s o3 report a mixed picture: Grok 3 slightly leads on some math benchmarks (e.g., AIME 2025 under heavy consensus sampling), while o3 (and related OpenAI reasoning models) lead on others, such as Codeforces coding Elo and certain software‑engineering tasks. These articles also note concerns that xAI’s benchmark setups (e.g., very expensive consensus sampling for Grok 3) aren’t perfectly comparable to how OpenAI models are typically evaluated, making it hard to declare an overall winner. (portotheme.com)

Why this is rated ambiguous The prediction effectively claims that by January/February 2025 Grok 3 would surpass OpenAI/Microsoft’s best publicly available frontier model and be the state‑of‑the‑art general‑purpose LLM. By late February 2025, Grok 3:

• Was clearly a top‑tier frontier model and #1 on one prominent user‑preference leaderboard (Chatbot Arena). (twitter.com)
• But faced credible, well‑publicized expert assessments and technical comparisons saying it did not clearly surpass OpenAI’s leading models overall.

Because “state‑of‑the‑art general‑purpose LLM” is not defined by a single universally accepted metric, and high‑quality sources disagree—some framing Grok 3 as SOTA and others explicitly saying OpenAI remained ahead or at least not clearly behind—the outcome of the prediction cannot be determined in a definitive, objective way. Hence the result is best characterized as ambiguous, rather than clearly right or clearly wrong.

The prediction is about the future ("in 2 or 3 years" from a December 7, 2024 podcast, i.e., roughly 2026–2027). As of today (November 30, 2025), those years have not yet occurred, so we cannot definitively say whether a single top-tier frontier AI model will cost on the order of US$100 billion to train.

Public reporting to date on frontier model training costs (e.g., estimates for models like GPT‑4 and successor projects) generally puts total training expenditures in the range of hundreds of millions to low single-digit billions of dollars per model—not anywhere near $100 billion yet. These figures are based on industry analyses and journalistic estimates of GPU usage, cloud spending, and datacenter investments, and even very large announced AI infrastructure projects (e.g., multi‑tens‑of‑billions datacenters) are spread across multiple years and many models, not a single training run. However, these current numbers do not resolve whether by 2026–2027 costs for a single state‑of‑the‑art model will reach ~$100B.

Because the time window for the prediction has not arrived and the cost level it refers to has not yet been (dis)proven by events in those years, the correct classification is: it is too early to tell.

As of late 2025, surveys and industry data show that natural‑language prompting is a widely used assistant, not the primary interface for most developers, so Friedberg’s claim that “human language will be the dominant programming language” in 2025 has not materialized. Stack Overflow’s 2025 Developer Survey reports that 84% of respondents use or plan to use AI tools, but many use them only occasionally, 46% explicitly distrust AI output, and about three‑quarters still prefer asking human colleagues over AI for help, indicating that AI (and thus natural‑language prompts) is supplementary rather than central. (techradar.com) Studies of GitHub Copilot and large enterprise deployments find that roughly 20–33% of accepted code lines come from AI and that even at Microsoft, AI contributes only around 30% of code in some projects—meaning most code is still written and edited directly in conventional languages like Python and JavaScript. (arxiv.org) While 2025 has seen the rise of “vibe coding” and rhetoric that human language is becoming a new kind of programming language, these are consistently described as emerging trends, used mainly for prototyping or specific workflows, and accompanied by significant concerns around security and reliability, not as the default way the majority of professional software is produced. (rollingai.news) Some industry analyses claim that around 41% of global code in 2024 was AI‑generated and that a large share of vibe‑coding users are non‑developers, but even these optimistic estimates imply that most code is still authored through traditional programming and apply to subsets of teams rather than “most developers” worldwide. (secondtalent.com) Overall, natural‑language interfaces have become important and fast‑growing in 2025, but the best available evidence shows they augment rather than replace conventional programming languages, so the prediction is best judged as wrong.

Donald Trump was inaugurated for his second (non‑consecutive) term as the 47th U.S. president on January 20, 2025. (en.wikipedia.org) The prediction explicitly concerns what the federal government will look like within 2–4 years of his taking office, so the relevant evaluation window runs roughly from January 20, 2027 to January 20, 2029.

As of the current date, November 30, 2025, less than one year of that term has elapsed. While Trump has already taken notable steps aimed at reshaping the federal bureaucracy—such as reinstating and rebranding the former “Schedule F” civil‑service category as “Schedule Policy/Career,” imposing a federal hiring freeze, ordering a full‑time return to in‑person work, and creating a new Department of Government Efficiency to oversee changes across agencies (en.wikipedia.org)—the prediction is about the recognizable end‑state of the government after 2–4 years, not just early moves.

Because the 2–4 year period the predictor specified has not yet passed, we cannot determine whether the federal government will ultimately "look very different" by the end of that window. The outcome could still plausibly validate or falsify the prediction, so at this time it remains too early to judge.

The prediction explicitly covers "the next few years" after the November 2024 election, i.e., roughly through 2028. As of November 30, 2025, only about one year has elapsed, and U.S. party platforms for 2026–2028 (including the 2028 Democratic platform) have not yet fully formed, been adopted, and played out in national elections.

While commentators do debate whether Democrats are moving left, right, or tacking to the center in response to Donald Trump’s 2024 victory, these are early, contested interpretations—not settled outcomes. The Democratic Party’s official 2028 platform does not exist yet, and any durable strategic repositioning toward the center (or away from it) can’t be conclusively assessed this far ahead of the prediction’s own deadline.

Because the prediction’s timeframe ("by around 2028") has not yet expired, there is not enough information to say it was right or wrong.

As of November 30, 2025, the U.S. dollar remains the dominant global reserve and settlement currency, and no single credible alternative reserve asset associated with BRICS (or otherwise) has actually displaced it in a structural way.

Key points:

1. USD still dominant as reserve currency

   • IMF COFER data and BIS/IMF commentary show the U.S. dollar continues to account for roughly 55–60% of disclosed global foreign‑exchange reserves; the euro, yen, pound, and others make up most of the rest. There is some gradual diversification, but nothing close to a replacement of the dollar as the primary reserve asset.

2. BRICS currency / settlement efforts still nascent

   • BRICS countries have discussed and, in some cases, piloted alternatives: greater use of local currencies in trade settlements, expansion of the New Development Bank, and repeated talk of a “BRICS currency” or a commodity‑linked unit. However, as of late 2025 there is no widely adopted, liquid, reserve‑quality BRICS currency or asset that central banks hold at scale comparable to the dollar. These efforts are still mostly at the proposal or limited‑pilot stage, not a mature global reserve alternative.

3. Shift into gold and Bitcoin is noticeable but not of the described type

   • Central banks—especially in emerging markets—have indeed been increasing gold purchases in recent years, partly as a diversification hedge. Bitcoin and other cryptoassets have seen cycles of speculative inflows and price appreciation, but they are not being adopted as primary reserve assets by central banks in a way that constitutes a systemic alternative to the dollar.

4. Timeline has not expired yet

   • The normalized prediction explicitly allows "within the next few years (by the end of 2027)." It is currently 2025, so there are more than two years left for this to play out. Even though we can say the prediction is not true yet, it is premature to label it definitively right or wrong.

Because the deadline is in the future and the required outcome (a credible alternative reserve asset significantly shifting global reserves and settlements away from USD) has not clearly materialized so far, the correct status is “inconclusive (too early)”, not right or wrong.

Based on data through late 2025, the prediction has not come true yet, but its timeframe (“at some point in the coming years”) is long enough that it can’t be definitively judged.

1. Fed Treasury purchases have not significantly increased; they’ve declined.
   Federal Reserve balance sheet reports show that from September 25, 2024 to March 26, 2025, total assets fell from about $7.1T to $6.7T, with securities held outright down $240B and U.S. Treasury holdings down from about $4.38T to $4.24T. (federalreserve.gov) Subsequent H.4.1 data for October 1, 2025 show Reserve Bank credit and securities held outright continuing to fall, with U.S. Treasury securities down to about $4.20T, confirming ongoing balance sheet runoff (quantitative tightening), not renewed large-scale buying. (federalreserve.gov) A Congressional Research Service summary for March 26, 2025 likewise notes about $4.2T in Treasuries on the Fed’s balance sheet, consistent with a smaller, not larger, Fed footprint. (congress.gov) This is the opposite of “needing to buy the debt” in a new monetization phase.

2. There is still substantial private and foreign demand for Treasuries.
   Market coverage of a 2025 30‑year Treasury auction reports “strong demand,” with the issue clearing at a yield below pre‑auction levels and a higher‑than‑average combined share for direct and indirect bidders, suggesting that investors remain willing to absorb new supply. (barrons.com) Broader commentary notes that despite de‑dollarization talk and some reduction in foreign central bank custody holdings at the New York Fed, total foreign holdings of Treasuries reached a record around March 2025, implying ongoing foreign (especially private-sector) appetite for U.S. debt. (reuters.com) That contradicts the near‑term premise that “there is no one else to buy the debt.”

3. But the prediction’s horizon is multi‑year, so it can’t yet be falsified.
   The statement was that “at some point in the coming years” the Fed will be forced to monetize federal debt because other buyers will be insufficient. Only a bit more than one year has elapsed since the October 25, 2024 episode, and current evidence only shows that this scenario has not occurred yet—not that it cannot occur later in the remaining “coming years.”

Because (a) the Fed has not significantly ramped up Treasury purchases and demand from other buyers remains adequate so far, but (b) the time window of the prediction clearly extends beyond late 2025, the correct classification at this point is inconclusive (too early) rather than right or wrong.

Key elements of Friedberg’s thesis have partially materialized, but the full commodity/inflation cycle he was talking about is still playing out, and performance across commodity‑linked businesses is very uneven.

1. Did the “upcoming inflationary environment with rising commodity prices” occur?

• Global inflation has eased from 2022 peaks but remains above pre‑COVID norms, with the IMF projecting global headline inflation around 4.2% in 2025, and U.S. inflation remaining above target with upside risks. This is described as persistent and more stubborn than hoped. (meetings.imf.org)
• U.S. data show renewed price pressures in 2025 (e.g., September 2025 CPI and PPI up 0.3% month‑on‑month, with producer prices driven by higher energy costs), and forecasters expect tariffs and input costs to keep inflation elevated. (reuters.com)
• Commodity prices are mixed: the Bloomberg Commodity Index is up about 15% over the last year as of November 26, 2025, indicating a broad rebound. (ycharts.com) However, the World Bank’s April and October 2025 Commodity Markets Outlooks project overall commodity prices to fall 7–12% in 2025 and further in 2026, driven mainly by a significant oil glut and weaker global growth. (worldbank.org)
• Within that aggregate, precious metals have exploded upward: gold and silver hit all‑time highs in October 2025 (gold briefly around $4,300–$4,400/oz; silver above $54/oz), driven by inflation and “currency debasement” fears. (markets.chroniclejournal.com)

Net: we do have an inflationary backdrop with high and sticky inflation and some strongly rising commodities (especially precious metals), but not a clean, broad‑based commodity bull market—energy and some bulk commodities are weak.

2. Have commodity‑linked businesses outperformed non‑commodity businesses “over that cycle”?
Evidence is mixed and depends heavily on which commodity segment you look at and what horizon you choose:

• Metals & mining equities:

  • The FTSE Global All Cap Precious Metals & Mining Index is up ~85.7% year‑to‑date in 2025 through August 29, 2025, outperforming the broad FTSE Global All Cap equity index (up ~14.75% YTD) by more than 70 percentage points. (lseg.com)
  • The MSCI World Metals & Mining Index shows a one‑year total return of ~27.9% as of November 21, 2025, well above global equity benchmarks, which are in the high‑teens to low‑20s over similar 12‑month windows (e.g., MSCI ACWI IMI ~17.3% 1‑year as of September 30, 2025). (ycharts.com)
    This is strong evidence that at least some mining/commodity‑linked stocks have substantially outperformed broad non‑commodity equities during the recent inflationary episode.

• Energy equities:

  • The S&P 500 Energy sector has a negative ~‑3.7% 1‑year total return as of November 25, 2025, badly lagging the S&P 500, which is up roughly mid‑teens percent in 2025. (ycharts.com)
  • World Bank and other analyses attribute much of this underperformance to an oil supply glut and expected further declines in energy prices through 2026. (worldbank.org)
    So energy producers—also quintessential “commodity‑linked businesses”—have not outperformed the broader market.

• Commodity trading houses:

  • Physical commodity traders (Vitol, Trafigura, Mercuria, Gunvor, Glencore, etc.) enjoyed record profits in 2022–23, then saw earnings fall sharply in 2024 as volatility normalized, though profits remained historically high. (spglobal.com)
  • By 2025, reports describe trading margins as having “normalized” from boom levels, with some rebound in metals trading in particular, but not a clear, broad step‑change that would let us say they have decisively beaten non‑commodity sectors through a full new cycle. (markets.financialcontent.com)

Given this:

• Certain commodity‑linked segments (especially precious metals and some metals/miners) are dramatically outperforming broad equity indices in 2025. (lseg.com)
• Other major commodity‑linked sectors (notably energy producers) are clearly underperforming. (ycharts.com)
• The “commodity trading businesses” Friedberg mentioned have moved from exceptional boom profits to more normal—though still high—levels, not a simple, sustained step‑up versus the rest of corporate earnings. (spglobal.com)

3. Is the cycle far enough along to judge his claim “over that cycle”?

• The podcast was released October 25, 2024, so only about 13 months have passed.
• Macro forecasts from the IMF and World Bank see inflation gradually declining but still above target in some advanced economies (especially the U.S.) through at least 2026, and they project commodity prices in aggregate to decline further into 2026, not obviously completing a new, clean commodity “up cycle” yet. (worldbank.org)
• Because Friedberg’s statement is explicitly conditional and multi‑year (“over that cycle”), the relevant comparison period is the full inflation/commodity cycle he anticipated, not just the first year of it. Right now we’re mid‑stream, and forward‑looking projections are still being revised.

Conclusion:

• We can say that parts of his thesis look directionally plausible: in a period of renewed inflation worries and strong precious‑metals rallies, some commodity‑linked equities (particularly gold/silver miners and broader metals & mining indices) have massively outperformed broad global equities so far. (lseg.com)
• However, other major commodity‑linked sectors (like oil & gas producers) have underperformed, and the aggregate commodity price outlook is for declines into 2026 rather than a clear, sustained up‑cycle. (ycharts.com)
• Because the time horizon and exact definition of the “cycle” are vague, and we are only about a year past the prediction, there isn’t yet a clear, cycle‑complete data set that would definitively show that “commodity‑linked businesses” as a whole out‑ or under‑performed non‑commodity businesses across the entire episode.

For those reasons, the fairest assessment today is “inconclusive (too early)” rather than clearly right or wrong.

As of November 30, 2025, the conditions Friedberg specified for the prediction have not occurred, and his "next couple of years" timeline from October 18, 2024 has not yet elapsed.

1. No “routine” 10× reuse of Super Heavy boosters yet
   Public flight records show that Starship has conducted about ten–eleven integrated test flights so far. Some Super Heavy boosters have been reflown (e.g., the booster for Flight 9 previously flew on Flight 7; the booster for Flight 11 previously flew on Flight 8), but no Super Heavy has anything close to 10 flights yet. (supercluster.com)

2. No hour‑scale turnaround; cadence is still weeks to months
   Launch-tracking data list the fastest turnaround between Starship system flights at roughly 37 days, not ~1 hour, and each individual booster/ship pair is used only occasionally, with long periods of inspection and modification between flights. (nextspaceflight.com)

3. $10/kg to LEO is still aspirational, not realized
   Current analyses of early Starship operations suggest effective launch costs on the order of tens to hundreds of dollars per kilogram—e.g., modeling around ~$90–100/kg with modest reuse—and industry writeups talk about projected or long‑term possibilities for $10–20/kg, not actual achieved prices. (nextbigfuture.com)
   Musk and commentators explicitly describe the ~$10/kg number as a future goal dependent on very high reusability and mass production, not something SpaceX is charging or achieving in 2025. (nextbigfuture.com)

Because:

• The prerequisite state Friedberg described (routine ~10‑flight reuse with ~1‑hour turnaround) has not been reached; and
• His timeline of "over the next couple of years" from October 2024 extends to roughly late 2026,

we cannot yet say whether his conditional claim about eventual $10/kg under those conditions is right or wrong. The correct status as of now is that it remains untested and too early to judge.

The prediction is about long‑term potential, not about reaching 100M+ subscribers by a specific year.

1. Current Starlink scale vs. the prediction

   • SpaceX confirmed Starlink passed about 4 million customers in late September 2024, matching the baseline figure discussed on the podcast. (techcrunch.com)
   • Public data and reporting show growth to roughly 8–8.5 million users by late 2025, far short of the 100M+ range discussed in the prediction. (en.wikipedia.org)

2. Condition on satellite‑to‑phone availability
   The prediction explicitly hinges on “if we can get satellite to phone and you can get Starlink more broadly available”. As of late 2025, Starlink has begun satellite‑to‑phone services (e.g., text messaging in partnership with mobile carriers) but does not yet have fully broad, mature satellite‑to‑phone coverage for mass‑market data and voice; this is still being built out via spectrum deals and new satellite technology. (washingtonpost.com)

3. No clear falsification window
   The claim is framed as “over time… this could be a 100 million subscriber business… it could be the largest subscription business in the history of humanity.” Since no time horizon is specified and the key prerequisite (ubiquitous satellite‑to‑phone) is still emerging, we cannot yet say whether Starlink will or will not eventually reach 100M+ subscribers and become the largest subscription business by count.

Because the claim is explicitly conditional and open‑ended in time, and Starlink is still far from 100M subscribers but growing quickly, the correct evaluation as of November 30, 2025 is “inconclusive (too early to tell).”

The prediction is explicitly about what will be needed over the next 10–20 years starting from 2024: that nuclear (uranium-based) power will have to provide a substantial share of new generation capacity to meet projected demand, and that renewables alone won’t be sufficient.

That time window runs roughly from 2024 to 2034–2044, so as of the current date (2025-11-30) we are only about one year into the 10–20 year horizon. It is far too early to assess whether, by the end of that period, actual policy, build-out, and grid mix will or will not conform to this claim.

Because the prediction is about a structural outcome over a decade or more, rather than an event on a specific near-term date, there is not yet enough information to determine correctness either way. Therefore the status is inconclusive (too early to tell).

The prediction is explicitly framed as playing out "over the coming years and decades," so only about one year of that horizon has elapsed since the October 18, 2024 podcast—far too little time to judge a decades‑scale forecast.

Early evidence does support parts of Friedberg’s premise: China is aggressively expanding nuclear capacity (over 100 reactors operational/under construction/approved and plans for many more, including advanced designs), putting it on track to surpass the U.S. in nuclear generation in the 2030s. (chinadailyhk.com) Meanwhile, U.S. deployment of new reactors has been slower, and analysts often describe China as 10–15 years ahead in nuclear build‑out. (itif.org) However, the U.S. is also moving to accelerate advanced nuclear (e.g., recent White House actions, microreactor programs, and plans for new AP1000 fleets), which cuts against a clear "Luddite state" narrative at this stage. (whitehouse.gov)

Because the core claim is about what "most of the world" and the U.S. will look like many years from now, and current signals could still reverse or accelerate, the forecast cannot yet be classified as definitively right or wrong.

The prediction’s explicit horizon is “over the next decades” and refers to structural, long‑run shifts in the global economy as it becomes more digital and automated. As of November 30, 2025, only about one year has elapsed since the podcast (October 18, 2024), whereas “decades” implies something like 20–30+ years (e.g., into the 2040s–2050s). There is not nearly enough time elapsed to judge:

• whether countries with cheaper, more abundant electricity systematically achieve superior economic performance in a clearly attributable way, and
• whether the United States suffers negative economic consequences relative to competitors due specifically to being slower in achieving low‑cost, abundant power.

These are long‑term comparative outcomes that can’t be reliably evaluated on a ~1‑year timescale, regardless of any short‑term data on energy prices, deployment of renewables or nuclear, or GDP growth. Therefore, the correct classification today is “inconclusive (too early)”.

The prediction is conditional and medium‑term: it says that if China can get large‑scale electricity down to about $0.05/kWh while U.S. power stays around $0.20/kWh, the U.S. will lose economic competitiveness relative to China.

As of late 2025, that core price-gap condition has not been met at the economy‑wide level:

• Recent cross‑country comparisons put average industrial electricity prices around $0.10/kWh in China and $0.08/kWh in the U.S. in 2025, i.e., the U.S. is not at $0.20 and is actually slightly cheaper for industrial users on average. (rhnuttall.co.uk)
• Detailed Chinese tariff tables show typical industrial non–time-of-use prices in major provinces in the 0.65–0.81 RMB/kWh range (roughly $0.09–$0.11/kWh at current exchange rates), again far above $0.05/kWh for standard industrial loads. (china-briefing.com)
• U.S. data for 2024–2025 show national retail averages roughly in the $0.13–$0.19/kWh band for commercial/residential users, while industrial averages are closer to $0.08–$0.09/kWh, not $0.20/kWh. (uschamber.com)

There are narrow cases in China where subsidized rates for AI data centers approach the posited level: several inland provinces offer a 50% discount on industrial tariffs, bringing power for qualifying AI data centers to about 0.4 RMB/kWh (~$0.056/kWh). (tomshardware.com) But this is:

1. Targeted to a specific sector and region, not the whole economy; and
2. Not paired with a persistent, economy‑wide U.S. price of ~$0.20/kWh for comparable large‑scale users (U.S. wholesale prices are closer to $0.04/kWh, and large industrial customers typically negotiate single‑digit‑cent tariffs). (reuters.com)

Because the hypothetical price configuration (China ≈$0.05, U.S. ≈$0.20 for large‑scale users) has not actually occurred across the two economies, we cannot empirically test the second part of the prediction (that the U.S. would then lose economic competitiveness). Moreover, Friedberg’s original wording was more of a counterfactual warning than a time‑stamped forecast that this configuration would definitely happen by 2025.

Given that:

• The relevant condition has not clearly materialized; and
• The time horizon he referred to (“over the coming years”) extends beyond 2025,

it’s too early, and the scenario too unmet, to label the prediction as right or wrong. Hence the status is best classified as "inconclusive".

The prediction was that, following the U.S. Supreme Court’s Chevron/Loper Bright decision, the California Coastal Commission’s scope and authority would be ‘peeled back’ and ‘dialed down’ over the ensuing years.

In June 2024, the Supreme Court decided Loper Bright Enterprises v. Raimondo, expressly overruling Chevron and ending the federal doctrine of automatic judicial deference to agency interpretations of ambiguous statutes, reinforcing a broader anti‑regulatory legal environment. (mayerbrown.com) Although that ruling formally concerns federal, not state, agencies, the prediction was about downstream political and regulatory effects on the California Coastal Commission.

Since the podcast (October 2024), there have in fact been concrete legal and regulatory moves that reduce the Commission’s authority in important contexts. In January 2025, Governor Gavin Newsom declared a wildfire state of emergency and issued Executive Order N‑4‑25, using his emergency powers under Government Code section 8571 to suspend CEQA and all California Coastal Act permitting requirements for rebuilding structures destroyed or damaged by the Los Angeles and Ventura County fires. (ceqadevelopments.com) This order temporarily removes the Commission’s permitting jurisdiction over a large set of coastal rebuild projects.

When the Commission issued guidance attempting to continue applying statutory Coastal Act exemption criteria to those projects, Newsom responded with Executive Order N‑14‑25 (January 27, 2025), explicitly clarifying that all Coastal Act requirements, including those exemption provisions, were suspended and directing the Commission not to issue guidance or take any action suggesting that Coastal Act permitting procedures still applied to covered rebuilds. (dailyjournal.com) Environmental and legal analyses summarize that this series of executive orders (N‑2‑25, N‑4‑25, N‑9‑25, N‑14‑25, later N‑20‑25) effectively removes state coastal permitting requirements for wildfire rebuilding in the affected coastal areas, with Commission staff acknowledging that the Coastal Act will not impede those rebuilds. (la.surfrider.org) This is a clear, formal dialing‑down of the Commission’s authority in a substantial portion of the coastal zone, implemented through regulatory and executive changes after Loper Bright.

Parallel political efforts have also targeted the Commission. Representative Kevin Kiley introduced the Coastal Commission Accountability Act, which would require California to automatically comply with federal coastal decisions, stripping the Commission of much of its autonomous regulatory power under the Coastal Zone Management Act; while this federal bill has not (yet) been enacted, it illustrates ongoing, post‑Chevron attacks on the Commission’s authority. (sfgate.com) At the same time, some litigation has upheld the Commission’s powers (for example, a 2025 tentative ruling in Sable Offshore’s Santa Ynez pipeline dispute that sustained the Commission’s cease‑and‑desist order and rejected the company’s challenge to the agency’s authority), showing that its core legal framework remains intact. (reuters.com)

Netting this out, however, the prediction was not that the Commission would be abolished, but that its scope and authority would be reduced in the years following the Chevron/Loper Bright decision. The post‑2024 emergency executive orders suspending Coastal Act permitting and ordering the Commission to stand down in large wildfire‑affected coastal areas are significant, real‑world examples of its authority being ‘peeled back’ and ‘dialed down’ via legal and regulatory action in that period. Although the changes are targeted and partly temporary, they align with the substance and timing of the forecast. On balance, the prediction is best classified as right, with the caveat that the long‑term, statewide extent of any structural weakening is still evolving.

Evidence as of late 2025 shows that the 2020 IMO sulfur cap has significantly increased radiative forcing over the oceans and likely accelerated recent warming, but it is too early and too uncertain to say that the rate of ocean warming in the 2020s (and beyond) is roughly double its pre‑2020 value.

Key points:

• The 2020 IMO regulation cut the sulfur content of marine fuel from 3.5% to 0.5%, reducing shipping-related SO₂ emissions by ~60–80%, and thus reducing reflective sulfate aerosols over the oceans. This removes a cooling mask and adds positive radiative forcing.⁽¹⁾⁽²⁾
• A high-profile study by Yuan et al. (2024) estimates a radiative forcing of about +0.2 ± 0.11 W/m² averaged over the global ocean from the shipping sulfur cut, and explicitly states that this could lead to a doubling (or more) of the warming rate in the 2020s compared with the rate since 1980, with strong spatial variability. NASA’s own summary repeats that this forcing “could lead to a doubling of the warming rate in the 2020s.”⁽³⁾⁽⁴⁾ This is strongly consistent with Friedberg’s mechanism and order of magnitude, but still framed as a projection.
• Other recent modeling studies find much smaller effects. Yoshioka et al. (2024) estimate a global aerosol effective radiative forcing from IMO 2020 of only ~0.13 W/m², yielding ~0.04–0.05 °C of additional global warming over several decades—equivalent to just a few extra years of greenhouse‑gas warming, not a sustained doubling of the long‑term rate.⁽⁵⁾ Jordan et al. (2024) similarly conclude that IMO 2020 accelerates global warming by roughly 2–3 years, not by a factor of two in trend.⁽⁶⁾ A multi‑model intercomparison (Partanen et al. 2024) also reports relatively weak forcing, comparable to only a couple years of recent CO₂ increases.⁽⁷⁾ These results conflict with the much larger effect inferred by Yuan et al.
• Observationally, global ocean heat content has set new records in 2020–2024 and shows an ongoing multi‑decadal acceleration, with 2024 again the hottest ocean year on record.⁽⁸⁾ However, disentangling how much of the change in the rate of ocean warming since 2020 is due specifically to shipping aerosols versus greenhouse gases, El Niño, and other factors is still under active debate. Some prominent scientists (e.g., Michael Mann) argue that ocean heat content is increasing steadily but not accelerating as dramatically as claimed by studies like Hansen et al., underscoring current scientific disagreement on the magnitude of any recent rate jump.⁽⁹⁾
• Statistically, robustly establishing that a decadal warming rate has doubled requires analyzing the full decade with enough signal-to-noise to overcome internal climate variability. As of November 2025 we are only halfway through the 2020s, and the prediction also extends “and going forward,” which refers to time periods that have not yet occurred.

Because:

1. the main paper that directly supports a potential “doubling” frames it as could lead to, not as an already-demonstrated fact,⁽³⁾⁽⁴⁾
2. other state-of-the-art studies find a significantly smaller acceleration,⁽⁵⁾⁽⁶⁾⁽⁷⁾ and
3. only half the decade has elapsed, making trend detection and attribution inherently premature,

the claim that the rate of ocean warming in the 2020s and thereafter will be roughly twice the pre‑IMO‑2020 rate cannot yet be judged as clearly right or wrong. It remains scientifically inconclusive (too early).

Sources

1. IMO sulfur reductions and post‑2020 average fuel sulfur content changes.⁽²⁾
2. Yoshioka et al. 2024 on ~80% reduction in shipping sulfur emissions.⁽²⁾
3. Yuan et al. 2024, Communications Earth & Environment (NASA NTRS abstract): +0.2 ± 0.11 W/m² over global ocean, potentially doubling 2020s warming rate vs. since 1980.⁽³⁾
4. NASA 2024 Year-in-Review summary restating Yuan et al.’s “could lead to a doubling of the warming rate in the 2020s.”⁽⁴⁾
5. Yoshioka et al. 2024, Atmospheric Chemistry and Physics: ERF ≈ 0.13 W/m², ~0.04–0.05 °C extra warming over coming decades.⁽⁵⁾
6. Jordan et al. 2024, Earth’s Future: IMO 2020 regulations accelerate global warming by ~2–3 years (~0.05 °C between 2020 and 2029).⁽⁶⁾
7. Partanen et al. 2024, Atmospheric Chemistry and Physics: multi‑model ERF from IMO 2020 comparable to a few years of recent CO₂ increase.⁽⁷⁾
8. Recent ocean heat content syntheses showing record OHC in 2020–2024 and ongoing acceleration.⁽⁸⁾
9. Coverage of Hansen’s higher-acceleration claims and Mann’s critique that ocean heat content is not accelerating as sharply as suggested.⁽⁹⁾

Citations: (1)–(9) correspond respectively to: 2, 2, 3, 3, 2, 2, 2, 1, 3 in the tool output: ⁽¹⁾[turn2search5] ⁽²⁾[turn2search0] ⁽³⁾[turn3search0]/[turn2search3] ⁽⁴⁾[turn3search1] ⁽⁵⁾[turn2search0] ⁽⁶⁾[turn2search8] ⁽⁷⁾[turn2search7] ⁽⁸⁾[turn1search12] ⁽⁹⁾[turn3search4]

This prediction is about multi‑year to multi‑decadal changes in the frequency of rare storms (events with historical return periods of 1-in-100 or 1-in-500 years) driven by persistently warmer oceans. The podcast aired on 11 October 2024, and we only have about one year of subsequent data (through 30 November 2025), which is far too short to robustly evaluate a statement about how often such rare events will occur on average.

A few relevant points from the science:

• U.S. data do show a long‑term increase in heavy precipitation and extreme one‑day rainfall events over the past century, with especially strong increases since the late 20th century. This is physically linked to warmer oceans and a more moisture‑laden atmosphere, and is expected to continue. (epa.gov)
• The IPCC and U.S. National Climate Assessment project that rare heavy‑precipitation events become more frequent as global warming progresses, but the quantified increases are generally factors of ~1.3–2 (or a few times more frequent) for many heavy‑rain extremes over the 21st century, not an immediate shift to nationwide 1‑in‑100‑year storms happening every 2–3 years. (en.wikipedia.org)
• The Asheville flooding from Hurricane Helene in September 2024 has been characterized as an event with an annual recurrence interval greater than 1,000 years in parts of the southern Appalachians—i.e., extremely rare in the historical statistics. (content-drupal.climate.gov) One such extreme event soon after the prediction is directionally consistent with concerns about more severe storms, but a single event doesn’t establish a new long‑term average frequency.
• A 2023 analysis cited by the Washington Post found that about 20% of the U.S. has already seen roughly a fourfold increase in the likelihood of 100‑year rainfall (from ~1% to ~4% yearly probability), which would correspond to something like a 1‑in‑25‑year event on the old scale—substantial, but still not at the level of “every 2–3 years” nationally, and it is regional rather than U.S.-wide. (washingtonpost.com)
• By definition, estimating or re‑estimating 100‑year or 500‑year return periods requires many decades of observations and/or detailed modeling, and NOAA guidance emphasizes that these probabilities are statistical constructs evaluated over long periods, not something that can be judged from one or two years of weather. (nwrfc.noaa.gov)

Taken together: current science and observations are consistent with the direction of Friedberg’s concern (warmer oceans → more intense and more frequent extreme precipitation and some record‑breaking events). However, his specific quantitative claim—that across the United States, 1‑in‑100‑year storms will occur roughly every 2–3 years and 1‑in‑500‑year events every few years—cannot be confirmed or falsified from only one additional year of data, and existing studies do not yet document such an extreme nationwide frequency shift. Therefore, it is too early to say whether this prediction is right or wrong.

The prediction is explicitly time-bound to “by the end of 2025” (i.e., around December 31, 2025). Today’s date is November 30, 2025, so the specified time horizon has not yet elapsed. While we can look at current data on U.S. federal debt and interest outlays, we cannot yet definitively say what the totals will be by year‑end 2025, including any late‑year issuance, policy changes, or interest‑rate moves. Therefore, the accuracy of this prediction cannot yet be determined and is too early to call.

It is too early to determine whether this prediction is right or wrong.

The claim is explicitly about the world roughly 10 years after September 27, 2024 (around late 2034). Today’s date is November 30, 2025, so only a little over one year has passed. The scenario the predictor is talking about (late 2034) has not occurred yet, so we cannot judge outcomes that depend on that future date.

What we can say about the present is that smartphones are still the dominant personal computing interface:

• In 2025, ~59–64% of global web traffic comes from mobile devices (primarily smartphones), with desktop around 36–40%. (mynewitguys.com)
• Articles summarizing 2025 data repeatedly describe smartphones as the main way people access the internet, indicating they remain the central personal computing device. (techgaged.com)
• Major tech leaders (e.g., Sundar Pichai at Google and Andrew Bosworth at Meta) say smartphones will stay dominant for at least the next several years, even as AR/smart glasses and ambient computing evolve. (timesofindia.indiatimes.com)

However, these observations only tell us that the transition has not happened yet in 2025; they do not prove that it won’t happen by ~2034. Because the prediction’s target date is still almost nine years away, the correct evaluation status is "inconclusive" (too early) rather than "right" or "wrong."

The prediction concerns the state of personal computing and interface paradigms around ten years after the episode date of September 27, 2024, i.e., roughly September 2034. Today is November 30, 2025, so fewer than two years have passed since the prediction was made and about nine years remain before the stated horizon. It is therefore too early to determine whether traditional smartphones will have largely disappeared as the dominant personal computing device or whether browser-based interaction will have largely ceased to be primary by that time. Because the forecast is explicitly about a future state well beyond the current date, its correctness cannot yet be evaluated.

As of November 30, 2025, less than two years have passed since the prediction was made and roughly 19 years remain until the ~2044 horizon implied by “in 20 years.” The claim is specifically about what will predominate as the dominant form of human–computer interaction for the next generation (his children’s generation) around that time, not about the situation in the mid‑2020s.

Currently, traditional direct control methods—keyboards, touchscreens, mice, and manual UI navigation—still dominate mainstream computing on phones, PCs, and tablets, even though ambient and AI‑mediated interactions (voice assistants, context‑aware services, generative‑AI copilots, etc.) are growing. There is no way to definitively assess now whether these ambient interactions will become the dominant mode by ~2044.

Because the prediction is explicitly about a 20‑year future state and that date has not yet arrived, its accuracy cannot yet be judged.

The prediction is framed over the “coming years,” which implies a multi‑year structural shift rather than something that should be fully visible by late 2025, only about a year after the podcast.

There are early signs consistent with stronger demand for in‑person human‑service work:

• Recent U.S. data show that health care and social assistance plus leisure and hospitality accounted for more than 100% of net job gains over the latest year, with other sectors stagnating or shrinking, indicating rising relative demand for face‑to‑face services. (axios.com)
• BLS long‑term projections to 2033–2034 expect particularly fast growth in human‑service‑heavy categories such as home health and personal care aides (+20.7%), personal care and service occupations overall (+6.4%), and construction and extraction (+5.6%), all above the average for all occupations. (blog.dol.gov)

However, the specific claim is that as AI and automation advance, there will be a “burgeoning and blossoming” of salaries, availability, and demand for in‑person human services. On that causal and magnitude question, the evidence is not yet clear:

• Global AI labor analyses (e.g., PwC’s 2025 AI Jobs Barometer) find the strongest wage and productivity gains in AI‑exposed sectors like finance and ICT, while less AI‑exposed sectors such as hospitality and some personal services show lower productivity gains and are not (yet) singled out for exceptional wage acceleration. (pwc.com)
• BLS projections attribute much of the growth in care and service jobs to demographics (aging population, chronic conditions) rather than to AI‑driven substitution elsewhere, and we do not yet have several years of realized wage data isolating the AI effect in trades, hospitality, fitness instruction, tutoring, and similar roles. (bls.gov)

Because (1) the stated horizon is longer than the time that has elapsed, and (2) the available data do not yet clearly show a broad, AI‑driven surge in both pay and job counts across the full range of in‑person human‑service roles he described, it is too early to say whether the prediction is ultimately right or wrong. Hence, inconclusive (too early).

The prediction has two parts:

1. Short‑term escalation into Lebanon (by the time the show aired)
   This did occur. On 23–30 September 2024, Israel launched Operation Northern Arrows, carrying out more than 1,700 airstrikes across Lebanon—over a 125% increase in Israeli strikes compared with August—killing hundreds and displacing large numbers of civilians. This marked a sharp expansion from earlier, more limited cross‑border fire. (en.wikipedia.org) On 30 September/1 October 2024 Israel then began a ground invasion of southern Lebanon, formally described as the 2024 Israeli invasion of Lebanon, confirming that the incursion had indeed “gotten bigger” in the days around airing. (downloads.regulations.gov) So the near‑term escalation call was accurate.

2. Probability (>30–40%) of a full‑blown multinational Middle East war before the November 2024 US election
   Here the specific scenario he highlighted did not happen:

   • The Lebanon front evolved into a large Israel–Hezbollah war plus an Israeli invasion of Lebanon, but contemporaneous descriptions frame it as the 2024 Israeli invasion of Lebanon and the Israel–Hezbollah conflict—not as a multi‑state regional war. (en.wikipedia.org)
   • Although Iran did launch direct missile/drone attacks on Israel in 2024 and multiple Iran‑aligned militias (in Iraq, Syria, Yemen) were active, mainstream assessments consistently spoke of the risk of a broader regional war rather than a realized multi‑state war, and noted that these conflicts had not escalated into wider regional wars by late 2024. (streetinsider.com)
   • Instead of widening into a multinational war before the US election (5 November 2024), the Israel–Hezbollah war was halted shortly afterward by a US‑ and France‑mediated Israel–Lebanon ceasefire agreement signed 26 November and effective 27 November 2024, explicitly aimed at ending hostilities between Israel and Hezbollah/Lebanon. (en.wikipedia.org) No Arab state (e.g., Egypt, Jordan, Saudi Arabia) or Iran entered into open conventional war with Israel in that period, nor did the US.

Because the core forecasted outcome—“a full blown multinational war in the Middle East” triggered by the Israel–Lebanon front before the 2024 US election—did not occur, the prediction is judged wrong for scoring purposes, even though one component (rapid escalation of Israel’s Lebanon operation between recording and airing) was correct and his stated probability (30–40%) was not a categorical claim of certainty.

Evidence about All‑In Summit 2024 shows that the on‑stage interviews were released to YouTube and podcast feeds over an extended period after (and at the tail end of) the September 8–10, 2024 event, not all at once.

• The 2024 All‑In Summit itself ran September 8–10, 2024 at Royce Hall in Los Angeles. (en.wikipedia.org)
• The All‑In Podcast YouTube video “Elon Musk | All‑In Summit 2024” is recorded as published on September 10, 2024, while the summit was concluding. (glasp.co)
• Additional “All‑In Summit 2024” sessions then appeared on later dates: a Marc Benioff summit episode released September 15, 2024; a John Mearsheimer and Jeffrey Sachs summit episode (with the corresponding YouTube video noted as published on September 16, 2024); and a Thomas Laffont | All‑In Summit 2024 episode whose YouTube summary shows a September 27, 2024 date, with the podcast version dated September 26, 2024. (allin.onpodcastai.com)
• The rollout continued into October 2024, e.g., the Juan Carlos Izpisua Belmonte / Altos Labs summit talk promoted with an October 1, 2024 publication date. (rumble.com)
• Listener discussions on Reddit in mid‑ to late‑September 2024 explicitly mention that summit sessions were being posted on YouTube and that not all had yet appeared on audio platforms, implying an ongoing staggered release. (reddit.com)

Taken together, these dates show summit content was indeed rolled out over successive days and weeks following the event, matching Friedberg’s prediction that, as in prior years, material from the 2024 All‑In Summit would be published gradually rather than all at once.

Contemporaneous writeups of the CME FedWatch Tool for the week ending around August 23, 2024 show that futures markets were indeed pricing very high odds of substantial 2024 cuts:

• An India Infoline summary of CME FedWatch probabilities for the week to August 23, 2024 notes that by December 2024 there was a 100% implied probability of at least 75 bps of rate cuts and about a 64.3% probability of 100 bps of cuts.【1†turn1search0】
• A related piece summarizing CME FedWatch around mid‑August 2024 reports essentially the same structure: 100% probability of 75 bps and roughly 63% probability of 100 bps of cuts by December 2024.【1†turn1search2】

Friedberg’s claim that “trading markets … are now estimating a 100% chance of a three quarter of a percent rate cut by the end of 2024 and a 70% chance of a one point rate cut” is therefore an accurate, rounded description of those CME FedWatch probabilities (100% vs. 100%, and ~64% vs. “roughly 70%”).

Since the normalized prediction is explicitly about what interest rate futures markets were implying at that time, and that implication matches independent summaries of the CME FedWatch data, the statement is best classified as right.

Multiple independent snapshots of market pricing around August 23, 2024 show that futures / CME FedWatch probabilities for the September 18, 2024 FOMC meeting were roughly:

• ~70–76% probability of a 25 bp cut
• ~24–30% probability of a 50 bp cut
• Essentially 0% probability of no cut (all probability mass on some size of cut)

For example:

• FX Leaders, citing the CME FedWatch tool on August 23, 2024, reported a 75.5% probability of a 25 bp cut and 24.5% probability of a 50 bp cut for September. (fxleaders.com)
• VStar’s August 26, 2024 market overview similarly noted 75.5% odds of a 25 bp cut and 24.5% odds of a 50 bp cut at the September FOMC. (vstar.com)
• A Reuters piece on August 23, 2024 (via Business Recorder) said traders were pricing a 73% chance of a 25 bp cut and 27% chance of a 50 bp cut next month, with no mention of a meaningful “no cut” probability. (brecorder.com)
• Intercam’s August 23, 2024 note cited FedWatch showing 71.5% probability of a 25 bp cut in September, implying roughly the remaining ~28.5% on a 50 bp cut. (intercamblog.com)
• A recap of FedWatch probabilities for the week ending August 16, 2024 similarly showed 71.5% for a 25 bp cut and 28.5% for a 50 bp cut at the September 2024 meeting. (indiainfoline.com)

Friedberg’s description — “50 or 75% chance of a quarter point cut. 20% chance of a 50 basis point cut and then 6% chance of no cut” — is an approximate, verbal summary of this distribution. The core structure matches the data:

• He correctly characterizes a large majority (roughly three-quarters) probability on a 25 bp cut, consistent with the 71–76% range actually implied by FedWatch and futures pricing.
• He assigns a smaller but significant probability (~20%) to a 50 bp cut, which is close (within a few percentage points) to the observed ~24–30% range.
• He notes a small residual probability (~6%) of no cut, which overstates what FedWatch snapshots suggest (essentially 0%), but still correctly conveys that no cut was viewed as a low-probability tail outcome.

Because the prediction was explicitly about what prediction markets were implying at that time, not about which outcome would actually occur, it should be judged against contemporaneous market data rather than the eventual decision (which was a 50 bp cut on September 18, 2024). (reuters.com) On that basis, his qualitative breakdown (big probability on 25 bp cut, smaller on 50 bp, very little on no cut) is substantively accurate, with only modest numerical imprecision.

Given the available evidence, the statement is basically correct as an approximate description of late‑August 2024 prediction‑market probabilities, so it is best classified as "right".

As of November 30, 2025, Brian Niccol has been Starbucks CEO since September 9, 2024, so we are only a bit over one year into his tenure, within a loosely defined "first couple of years" window. (about.starbucks.com) Under Niccol, Starbucks has expanded use of AI tools like Green Dot Assist in about 35 pilot stores, plus "Smart Queue" and an AI-driven inventory-counting system, but these are software assistants layered on top of normal barista work, not robot-heavy or visibly highly automated drink-prep stores. (businessinsider.com) Starbucks’ main in-store automation platform, the Siren Craft System, was developed and rolled out broadly before Niccol arrived, and since he took over the company has publicly said it is scaling back reliance on automation, slowing or limiting Siren’s deployment and instead hiring thousands of additional baristas and reversing prior labor-cutting moves. (about.starbucks.com) Niccol and company representatives have repeatedly emphasized that Starbucks is “not near” any fully robotic staff model and that the goal is to bring more human partners back into stores, while the announced "coffeehouse of the future" prototypes scheduled to open in fiscal 2026 are described in terms of ambiance, seating, and lower build cost rather than as heavily automated or robotic cafes. (people.com) I could not find evidence that Starbucks has opened any experimental stores under Niccol that are noticeably more automated than typical locations in the sense of using robots or heavy automation for drink/food preparation; however, because the predicted time window extends roughly through the first two years of his tenure (into late 2026) and those future prototypes have not yet opened, it is still too early to say definitively that such stores will not appear. Therefore, the prediction is currently inconclusive (too early to judge).

Brian Niccol was named Starbucks chairman and CEO on 13 August 2024 and started in the role on 9 September 2024, so his first year runs through early September 2025. (about.starbucks.com) In October 2024, shortly after taking over, he publicly said Starbucks needed to simplify its overly complex menu as part of a turnaround. (theguardian.com) On the Q1 FY2025 earnings call in late January 2025 (about four months into his tenure), he announced a Back to Starbucks plan that included cutting roughly 30% of Starbucks food and beverage items and reducing customization complexity in order to streamline operations and focus on fewer, better offerings. (nypost.com) Subsequent coverage noted that trimming around 30% of the menu was a central plank of his strategy and was already being implemented within his first six months. (cnbc.com) By February–March 2025, Starbucks had begun removing specific items (such as 13 low-performing drinks effective 4 March 2025) under this initiative, a clear reduction in menu breadth and complexity well before September 2025. (nypost.com) Because these concrete menu simplifications occurred within Niccol’s first year as CEO, the prediction that his early tenure would feature a reduction in Starbucks menu complexity is borne out.

Friedberg’s normalized prediction was that there was a greater‑than‑not (>50%) probability the U.S. would enter a recession within about one year of the August 9, 2024 episode (by ~August 2025).

By that date, the commonly used recession criteria were not met:

1. Two consecutive quarters of negative real GDP did not occur. Real U.S. GDP grew at annualized rates of about 3.1% in Q3 2024 and 2.4% in Q4 2024, then fell 0.5% in Q1 2025 and rebounded 3.3% in Q2 2025. That is, only one quarter (Q1 2025) was negative, and it was followed by strong positive growth, so the “two negative quarters” rule-of-thumb for a recession was not satisfied.

   • Q3 2024: +3.1% real GDP growth.
   • Q4 2024: +2.4%.
   • Q1 2025: –0.5% (contraction), largely attributed to a surge of imports ahead of new tariffs.
   • Q2 2025: +3.3% rebound.

   (apnews.com)

2. No officially dated NBER recession in that window. The National Bureau of Economic Research’s Business Cycle Dating pages still report the most recent peak in U.S. economic activity as February 2020 and the trough as April 2020, with no later peaks or troughs announced as of late 2025. (nber.org) Since NBER is the standard arbiter of U.S. recessions and has not declared a new recession after the short 2020 downturn, there is no recognized recession starting between August 2024 and August 2025.

3. Broader data and commentary describe a slowdown, not a recession in that period. Full‑year 2024 U.S. real GDP growth is estimated around 2.8%, with projections around 2% for 2025, indicating continued (if slower) expansion rather than an outright contraction. (en.wikipedia.org) While some analysts and even an NBER research paper discussed elevated probabilities (e.g., a model‑based 71% chance the U.S. was in recession as of May 2025), these were probabilistic assessments and not confirmations that a recession had actually begun. (linkedin.com)

Given that (a) the two‑negative‑quarter GDP criterion was not met within the forecast window, and (b) the NBER has not subsequently dated any recession in that period, the U.S. did not enter a generally recognized recession between August 9, 2024 and August 9, 2025. Consequently, a forecast that the probability of such a recession was greater than 50% turned out to be wrong in outcome (the event did not occur in the stated timeframe), even if many contemporaneous forecasters viewed recession risk as elevated.

The prediction was conditional: if the U.S. entered a recession sometime between roughly August 2024 and August 2025, then the federal government would roll out significant countercyclical programs that kept equities rallying or at least “strong” despite that recession.

On the first part – whether a recession actually occurred in that window – the evidence is mixed and not cleanly resolved:

• The NBER-based recession indicator used by the St. Louis Fed remains at 0 (expansion) through at least October 2025, meaning no official post‑2020 recession has been dated yet. (fred.stlouisfed.org)
• Major forecasters (IMF, RBC, UBS, various banks) describe a sharp slowdown / soggy or near‑zero growth and elevated recession risk in 2025, but generally do not say a clear, broad U.S. recession is underway; some emphasize the distinction between a “growth recession” and a formally defined recession. (livemint.com)
• Academic work using unemployment/vacancy classifiers estimates around a 71% probability the U.S. was already in recession as of May 2025, but that is probabilistic model evidence, not a widely accepted determination. (arxiv.org)

On the second part – government support and equity performance – the picture is also not cleanly attributable to what Friedberg described:

• The U.S. did suffer a sharp stock market crash in early April 2025 tied to Trump’s broad “Liberation Day” tariff shock, but by late August 2025 the S&P 500 had rebounded and was up about 9–10% year‑to‑date, implying equities were again relatively strong. (en.wikipedia.org)
• Federal actions in this period were dominated by aggressive tariff policy, talk of future tax cuts, and assurances from Treasury and the Fed that the system was sound and that tools were available if needed; there is little evidence of large, new, clearly countercyclical market‑support programs of the kind seen in 2008 or 2020 rolled out specifically in response to an acknowledged recession. (ft.com)

Because (1) it is not clearly established that the U.S. was in a formal, broad‑based recession during the specified period, and (2) any subsequent equity strength cannot be unambiguously tied to distinct, recession‑response government programs, the prediction cannot be judged definitively true or false on the available record. Hence the outcome is best classified as ambiguous rather than right, wrong, or merely “too early.”

As of November 30, 2025, Wiz remains a privately held company pending acquisition by Alphabet/Google, and therefore does not have a standalone public market equity valuation or market capitalization.

Key facts:

• Wiz raised $1 billion at a $12 billion valuation in May 2024, one of the highest valuations for a cybersecurity startup at that time. (techcrunch.com)
• Google’s initial 2024 acquisition talks at around a $23 billion valuation fell through, with Wiz choosing to pursue an eventual IPO instead. (forbes.com)
• In March 2025, Alphabet agreed to acquire Wiz for about $32 billion in cash, Google’s largest acquisition ever and the biggest cybersecurity deal to date. (techcrunch.com)
• That deal is expected to close in 2026, at which point Wiz will be absorbed into Google Cloud rather than trading independently as a public company. (spglobal.com)

The prediction being evaluated is: If Wiz maintains its current high growth rate, its public market valuation could reach roughly $100 billion, comparable to Palo Alto Networks, within about two years of July 26, 2024 (i.e., by mid‑2026).

Because:

1. The time horizon (mid‑2026) has not yet elapsed as of November 30, 2025; and
2. Wiz has not become a public company with its own market cap, but instead is slated to be acquired and integrated into Alphabet at a ~$32B valuation,

we cannot determine whether Wiz would have reached a ~$100B public market cap by mid‑2026 had it remained independent and gone public. There is no observable public-market test of the prediction yet, and the deadline has not passed.

Therefore, the correct status for this prediction is “inconclusive (too early)”.

Evidence from the days immediately after Biden’s July 21, 2024 withdrawal shows that Democratic and anti-Trump voters rapidly consolidated behind Kamala Harris and that her head‑to‑head numbers against Trump improved relative to Biden’s.

• Rapid consolidation of ‘anyone but Trump’/Dem-aligned voters: A CNN poll conducted July 22–23 (immediately after Biden exited) found Trump 49% – Harris 46%, a smaller deficit than Biden’s earlier 6‑point gap with Trump among the same respondents. The article notes that Democratic and Democratic‑leaning voters were “broadly enthusiastic” about Harris and “willing to coalesce around her” as the new presumptive nominee; 76% of them said the party should nominate her. Harris held 95% of prior Biden supporters and picked up a share of voters who had previously said they’d support neither Biden nor Trump. (abc17news.com) This reflects exactly the kind of quick rallying of anti‑Trump voters Friedberg described.

• Polling performance strengthened vs. Biden: Aggregated national polling before Biden’s withdrawal (Biden vs. Trump) showed Trump leading by about 2.8 points on average (Trump 46.9% – Biden 44.1% across major averages through July 23). (en.wikipedia.org) In the immediate post‑withdrawal period, multiple national polls showed Harris vs. Trump essentially tied or with Harris slightly ahead: e.g., Reuters/Ipsos July 22–23 (Harris 44% – Trump 42%), New York Times/Siena July 22–24 (Trump 48% – Harris 46%), and Morning Consult July 22–24 (Harris 46% – Trump 45%). (en.wikipedia.org) These are all better margins than Biden’s recent matchups.

• Contemporaneous analyses explicitly say she’s doing better than Biden: Reuters reported that both a battleground‑state Emerson/The Hill series and a New York Times/Siena national poll showed Harris performing better against Trump than Biden had, and described Harris as “gaining momentum” after entering the race. (reuters.com) Politico’s write‑up of a July 23 NPR/PBS/Marist poll (Trump 46% – Harris 45%) emphasized that Harris was already neck‑and‑neck with Trump just a day after launching and that she had “quickly consolidated support within the Democratic Party, securing delegate pledges.” (politico.com) The Guardian similarly noted that Harris had narrowed Trump’s lead and reset the race after Biden dropped out. (theguardian.com)

Taken together, these data show that in the near term after Biden’s withdrawal, Democratic/anti‑Trump voters did rapidly rally to Harris, and her polling against Trump improved relative to Biden’s. That matches Friedberg’s prediction, so it should be judged right.

It’s too early to evaluate this prediction. All of the claims are about outcomes in 2050 and 2060 (U.S. capacity doubling to ~2 TW by 2050; China reaching ~8.7 TW total capacity with ~88% renewables by 2050; ~18% of China’s power from nuclear by 2060). The current date is November 30, 2025, which is about 25 years before 2050 and 35 years before 2060, so there is no way yet to determine whether those long‑term build‑out and mix targets will be met. Even though there are various current plans and scenarios for U.S. and Chinese power systems, those are themselves projections and are not evidence that the 2050/2060 outcomes have already occurred or are guaranteed. Therefore the prediction cannot currently be judged as right or wrong.

The prediction is framed on a multi‑decade horizon (a buildout toward ~300 reactors and ~500 GW by around mid‑century), so as of November 2025 it is far too early to determine whether those long‑run targets will be met or not.

However, all observable developments since the July 26, 2024 podcast are directionally consistent with the prediction rather than contradicting it:

• At the time of the quote, China had about 26 reactors under construction; Chinese and international industry data still show China leading the world in new builds, with around 27–30 reactors under construction by end‑2024 / mid‑2025, and total operating+under‑construction units exceeding 100. (chinadaily.com.cn)
• The Chinese State Council has continued to approve large batches of new reactors. In August 2024 it approved projects totaling 16 reactors, and in April 2025 another batch of 10 reactors—the fourth consecutive year with approvals for at least 10 new units—described as a “peak period of large‑scale construction.” (worldnuclearreport.org)
• Policy targets also support ongoing expansion: recent reports from the China Nuclear Energy Association and related coverage cite goals around 110 GW of operating nuclear capacity by 2030 and ~200 GW by 2040, more than tripling current capacity and keeping nuclear as a major growth sector. (ecns.cn)
• Academic and scenario analyses used in policy discussions outline 2050 nuclear capacity in the 300–500 GW range, depending on how fast renewables scale—numbers that match the “300 reactors / ~500 GW” order‑of‑magnitude Friedberg referenced. (drganghe.github.io)

So far, China has continued and accelerated its nuclear buildout in line with those projections, but whether it will ultimately reach ~300 reactors and ~500 GW by mid‑century cannot yet be known. Hence the only defensible verdict now is that the prediction’s long‑term outcome is still inconclusive.

The prediction’s timeline is “in the future (by around 2050)”, while today is 2025-11-30, so we are only ~1/2 to 1/3 of the way through the period Friedberg is talking about. It is therefore too early to judge the long‑run claim that China’s average retail electricity price will fall below $0.06/kWh and remain well below U.S. retail prices.

What we can say so far:

• Recent data (2023–2024) show China’s average end‑user electricity prices generally in the $0.07–0.09/kWh range (about 0.45–0.65 CNY/kWh), while U.S. residential prices are much higher, typically $0.15–0.20+/kWh, with commercial/industrial somewhat lower.
• China is indeed expanding both renewable capacity (especially solar and wind) and nuclear power, with very large build‑out plans through 2035–2050, which could exert downward pressure on average costs, but outcomes will depend on many uncertain factors (grid investments, subsidies, fuel prices, policy, etc.).

Because the key part of the prediction is explicitly about the future level by ~2050 and the relative comparison at that time, and we are far from that date, the correctness of the prediction cannot yet be determined.

Donald Trump did, in fact, win the 2024 U.S. presidential election. Official results show Trump/ Vance defeating Harris/Walz with 312 electoral votes to 226 and winning the popular vote with about 49.8% to 48.3%.(en.wikipedia.org)

Joe Biden subsequently did drop out of the race on July 21, 2024, and Kamala Harris became the Democratic nominee; Trump still won against her in November.(en.wikipedia.org)

So in the scenario that actually occurred (Biden drops out and is replaced by Harris), Friedberg’s effective prediction that “it’s over, Trump’s won” was borne out. The stronger claim that Trump would win even if Biden had stayed in is counterfactual and can’t be directly tested, but standard forecast scoring evaluates only the realized world. On that basis, the prediction is judged right.

As of November 30, 2025, there hasn’t been enough time to judge a decade-scale prediction about “at least the next several years.” The evidence so far is mixed:

1. Realized CPI since mid‑2024
   After the July 12, 2024 podcast, U.S. headline CPI year‑over‑year has mostly run in the mid‑2% to low‑3% range. Data compiled from BLS releases show YoY CPI of about 2.4–3.5% through late 2024 and 2.3–3.0% for 2025, e.g. 2.31–2.39% in March–April 2025, then drifting back up toward ~3% by September 2025. (officialdata.org) That is above the Fed’s 2% target but below a clear 3% regime, and we have only ~1½ years of post‑prediction data—not enough to say whether inflation has sustainably failed to return to ~2% over “several years.”

2. Market expectations are closer to ~2–2.5% than to 3%
   Market‑implied 5‑year breakeven inflation has been around 2.2–2.6% since mid‑2024 and is about 2.3% in late 2025, not 3%. (ycharts.com) That contradicts the specific claim that markets would “assume that we’re going to be at a 3% kind of inflation level for quite some time,” at least so far.

3. Economist forecasts
   A late‑2025 survey by the National Association for Business Economics, summarized by Reuters, projects U.S. inflation easing from about 2.9% at the end of 2025 to around 2.6% in 2026. (reuters.com) That is still above 2% but again notably below 3%, suggesting consensus expects gradual convergence toward the low‑2s rather than a stable 3% regime.

4. Current underlying inflation
   Some gauges, including the Fed’s preferred measure, are still around 3% in late 2025, above target, with tariffs and policy uncertainty keeping upside risks alive. (reuters.com) This lends partial support to the idea that inflation may run above 2% for a while, but the direction of travel is downward from the post‑pandemic highs.

Putting this together: realized inflation since the podcast has indeed stayed above 2% and often in the high‑2s, which is loosely consistent with a world of somewhat higher inflation. However, market pricing and professional forecasts do not clearly back the notion of a durable 3% regime, and the forecast horizon (“for at least the next several years / on the order of a decade”) extends well beyond the data currently available. Because we cannot yet know whether inflation will or will not sustainably return to ~2% over that full period, the prediction’s long‑run accuracy is too early to call, hence inconclusive.

By November 30, 2025 we are ~16 months after the July 12, 2024 podcast, so only part‑way through Friedberg’s stated 24–36 month horizon for “extraordinary ROI” on the AI infrastructure buildout. The core of the prediction is thus still within its own time window.

On the evidence side, there are strong positive signals but not conclusive proof:

• Vendors supplying AI infrastructure, especially Nvidia, are generating enormous, highly profitable data‑center revenues driven by AI demand (tens of billions per quarter with very high margins, plus large backlogs for AI GPUs). These numbers show that at least some parts of the AI infrastructure stack are producing very strong financial returns.【2search1】【2search5】
• Hyperscalers (Microsoft, Amazon, Alphabet, Meta) have pushed capex heavily into AI data centers and GPUs—analysts estimate hundreds of billions in AI‑related capex, with cloud and AI services revenues growing rapidly. However, this spending is often framed as a long‑term strategic bet, with revenue growth still catching up to capex growth.【1search0】【1search5】

At the same time, multiple broad surveys and analyst reports indicate that AI ROI is not yet clearly or uniformly “proven” across the wider economy:

• Gartner and CIO‑focused coverage describe 2024–2025 as a "build‑out" or foundation‑laying phase where CIOs are increasing AI budgets substantially, but they emphasize that the ROI of many AI projects remains uncertain even as spending accelerates.【1search1】【1search7】
• CloudZero’s 2025 “State of AI Costs” report finds that only about half of organizations say they can confidently evaluate AI ROI, and that a large share of AI cloud spend is wasted on idle or poorly optimized resources.【1search6】 A separate analysis similarly notes AI costs rising sharply while ROI remains unclear for many firms.【1search4】
• Credit and bond‑market commentary on hyperscaler AI capex highlights that investors are effectively betting that future AI revenues will justify massive, debt‑funded build‑outs, but explicitly warns that a slowdown in AI adoption could leave companies overextended—i.e., the economic payoff is still a forward‑looking assumption, not yet a settled fact.【1news14】

Putting this together:

• The 2–3 year horizon has not yet elapsed, so we cannot fairly judge the ultimate claim that the AI infrastructure buildout will be “extraordinarily” ROI‑positive by mid‑2026 to mid‑2027.
• Within the nearer “next couple of quarters” sub‑claim, we do see strong profitability for chip vendors and some cloud segments, but broad, demonstrable proof that the overall AI infrastructure buildout is economically justified for hyperscalers and enterprises is not yet clearly established in the aggregate data and commentary.

Given that the primary window of the prediction is still open and the available evidence is mixed and evolving, the fairest assessment as of November 30, 2025 is “inconclusive (too early)” rather than clearly right or clearly wrong.

Multiple authoritative sources show that Donald Trump won the 2024 U.S. presidential election and was elected to a second (non-consecutive) term. The certified results list Trump as receiving 312 electoral votes to Democrat Kamala Harris’s 226, as well as winning the national popular vote plurality.(en.wikipedia.org) Overviews of the election and of Harris’s campaign likewise state that Harris lost to Trump on November 5, 2024.(en.wikipedia.org) Since Trump was subsequently inaugurated again on January 20, 2025,(en.wikipedia.org) Friedberg’s June 2024 prediction that Trump was likely "going to win a second term" was borne out by events.

Donald Trump’s 2024 campaign trade platform centered on broad, across‑the‑board tariffs — notably a 10% (or higher) universal tariff on all imports and around 60% on Chinese goods. FactCheck.org and other analyses describe these as economy‑wide proposals, not narrow sectoral measures.(factcheck.org)

In office during his second term, Trump has substantially implemented that broad vision:

• In April 2025 he invoked emergency powers (IEEPA) to impose a 10% baseline tariff on almost all U.S. imports, plus higher “reciprocal” tariffs for dozens of countries. Reuters describes this as a “sweeping overhaul” of the global trading system, not a narrow, sector‑limited policy.(reuters.com)
• Reporting and tax advisories summarize the new regime as a 10% across‑the‑board baseline with elevated country‑specific rates (e.g., 20–46% on major partners, and much higher on China).(grantthornton.com)
• For China specifically, tariffs were pushed far beyond the original 60% idea: combined measures briefly raised effective rates to about 145% on Chinese imports before being partially reduced in the context of negotiations, still leaving very elevated duties in place.(en.wikipedia.org)

There was a temporary 90‑day pause on many of the new above‑10% “reciprocal” tariffs after a market crash, but the 10% universal tariff remained in effect, sector‑specific duties on items like autos, steel and aluminum continued, and many country‑specific hikes were later implemented anyway. Estimates put the average applied U.S. tariff rate around 27%, the highest in more than a century — again indicating a broad, not narrowly targeted, regime.(en.wikipedia.org)

Friedberg’s prediction was that if Trump pursued this agenda, economic advisers would force a major rethink so that any actual tariff program would be toned down and confined to “particular slices of the economy” rather than implemented broadly. In reality, Trump has enacted and largely maintained a sweeping, economy‑wide tariff structure with only limited carve‑outs and tactical pauses, and in some respects (e.g., China rates) has gone well beyond the original headline numbers.

Because the broad, across‑the‑economy tariff policy has in fact been implemented and sustained, rather than significantly scaled back to narrow sectors, the prediction is wrong.

Contemporaneous U.S. data for mid‑2024 do not support classifying the macro environment as stagflation under the usual definition of high inflation plus stagnant/low growth and typically elevated unemployment.(en.wikipedia.org)

Growth: According to the BEA, real GDP grew at a 1.4% annual rate in Q1 2024 and then accelerated to a 3.0% annual rate in Q2 2024 (second estimate).(bea.gov)(bea.gov) That Q2 pace is solidly above most estimates of U.S. trend growth and not consistent with “stagnant” or “1–2%” growth as in the normalized prediction.

Inflation: CPI inflation was above the Fed’s 2% target but moderating: the 12‑month CPI increase was 3.3% in May 2024 and 3.0% in June 2024.(fraser.stlouisfed.org)(fraser.stlouisfed.org) The Fed’s preferred PCE measure showed Q2 2024 inflation around the mid‑2% range, with core PCE about 2.8–2.9% annualized.(bea.gov)(bea.gov) That is mildly elevated but not the kind of persistently high inflation usually associated with stagflation.

Unemployment: The unemployment rate in June 2024 was 4.1%, up from the lows of 2022–23 but still historically low and far from the high‑unemployment environment typically seen in stagflation episodes.(cnbc.com)

Putting this together, mid‑2024 featured moderate, above‑trend growth, easing but still‑above‑target inflation, and relatively low unemployment, which does not meet standard criteria for stagflation. The more specific normalized prediction that real GDP growth would remain in the 1–2% range through Q2 2024 is also contradicted by the actual ~3% growth reading. Therefore, Friedberg’s claim that “we have stagflation right now” is best judged as wrong.

By November 30 2025, only about 1.5 years have passed since the June 2024 prediction, which explicitly referred to a timeframe of a couple of years (roughly mid‑2026). That window has not yet elapsed, so the outcome cannot be definitively judged. Current data suggests the market has not yet shifted in the way described: surveys show around 80% of organizations pay for subscriptions to tools like ChatGPT or Microsoft Copilot and 63% are using cloud AI APIs, while only 39% are using open‑source models on their own infrastructure and 27% are training proprietary in‑house LLMs. (siliconangle.com) Another survey cited by the Financial Times reports that only about one in eight commercial AI workloads run on open models, with most customers preferring paid, state‑of‑the‑art proprietary systems and even open‑model users typically mixing them with commercial ones. (ft.com) A separate study finds open models account for only about 20 percent of usage and 4 percent of revenue in the AI market despite being up to 84 percent cheaper to operate. (itpro.com) Enterprise spending on generative AI is still increasing overall, with a majority of organizations planning to boost LLM spending, and OpenAI and Google models remaining widely used. (electronicspecifier.com) Netskope telemetry likewise shows that enterprise AI use is dominated by SaaS gen‑AI apps such as ChatGPT, Gemini, and Copilot, along with managed cloud AI platforms, rather than internally hosted open‑source stacks replacing external services. (reddit.com) These indicators imply the prediction has not clearly come true yet, but because the stated timeframe runs to around mid‑2026, it is still too early to call it definitively right or wrong.

The prediction claimed that the 2024 Atlantic hurricane season (the season following the June 7, 2024 episode) would produce hurricanes that were, in aggregate, larger, stronger, and more frequent than in any previously recorded Atlantic season.

What actually happened in 2024:

• The 2024 Atlantic season had 18 named storms, 11 hurricanes, and 5 major hurricanes, with an Accumulated Cyclone Energy (ACE) of about 161.5. It was extremely active and the third‑costliest season on record, but not record‑breaking in overall activity or intensity metrics. (en.wikipedia.org)

Comparison with previous top seasons:

• 2005 had 28 named storms, 15 hurricanes, 7 major hurricanes, and ACE ~245.3, with multiple Category 5 storms; it remains one of the most intense and active seasons ever recorded. (en.wikipedia.org)
• 2020 had 30 named storms (record), 14 hurricanes, 7 major hurricanes, and ACE ~180.3; it is explicitly described as the most active Atlantic hurricane season on record by number of storms. (en.wikipedia.org)

Thus, while 2024 was well above average and very destructive, its storm counts and ACE were lower than those of 2005 and 2020 and did not set all‑time records for frequency or overall energy. It therefore did not produce hurricanes that were, in aggregate, larger/stronger/more frequent than any previously recorded season. The probabilistic forecast he referenced did not “play out” in that strongest possible sense, so the normalized prediction is wrong.

Evidence from NOAA and related climate analyses shows that the broad conditional setup Friedberg described largely occurred, and the main real‑world consequence he warned about did materialize.

1. Sea surface temperatures in the Atlantic/MDR

   • NOAA and other analyses note that the Atlantic Main Development Region (MDR) was at record‑warm levels going into the 2024 season, with SSTs in May–June already comparable to, or above, what is normally seen near the peak of the season. (climate.gov)
   • CSU’s post‑season assessment attributes the 2024 hurricane environment in August–October to “extremely warm Atlantic sea surface temperatures” combined with cool‑neutral ENSO, producing record‑low vertical wind shear and extremely favorable late‑season conditions. (newsmediarelations.colostate.edu)
   • However, a detailed climate review for August 2024 reports that the MDR had its second‑warmest August on record, not the absolute warmest. (yaleclimateconnections.org) This means Friedberg’s phrase “record sea surface temperatures in the August to October time frame” is slightly overstated if interpreted as “calendar‑month all‑time record” for August, though daily and sub‑regional SSTs were still near record and exceptionally high.

2. Multiple massive Atlantic hurricanes making U.S. landfall in 2024

   • The 2024 Atlantic hurricane season was “extremely active and destructive,” with 18 named storms, 11 hurricanes, and 5 major hurricanes; NOAA and the NHC classify it as the third‑costliest Atlantic season on record. (en.wikipedia.org)
   • NOAA’s season wrap‑up notes that five hurricanes made landfall on the continental U.S., two of them as major hurricanes. (nesdis.noaa.gov) Those included:
     • Hurricane Helene (Category 4 at Florida landfall), which devastated the Big Bend region and parts of the Southeast and is described as one of the costliest and deadliest U.S. hurricanes on record. (en.wikipedia.org)
     • Hurricane Milton (Category 5 over the Gulf; Category 3 at Florida landfall), with over $34 billion in damage and widespread destruction in Florida and neighboring states. (en.wikipedia.org)
     • Additional U.S. hurricane landfalls included Beryl (Texas, Category 1), Debby (Florida, Category 1), and Francine (Louisiana, Category 2), all of which caused significant flooding and damage. (en.wikipedia.org)
   • CSU explicitly links the late‑season hyperactivity and these intense storms to the extremely warm Atlantic SSTs and low shear environment during August–October 2024. (newsmediarelations.colostate.edu)

Bottom line:
The Atlantic did remain anomalously and in many respects record‑warm into the peak of the 2024 season, and this environment did in fact produce multiple massive hurricanes from the Atlantic that struck the continental U.S. in 2024, notably Helene and Milton, plus several other hurricane landfalls. While the “record” wording for August MDR SST is not strictly perfect (it was second‑warmest by at least one metric), the core prediction—that continued extraordinary Atlantic warmth would yield multiple major U.S. hurricane impacts in 2024—was borne out. Hence the prediction is best judged as right, with a minor caveat on the precise SST superlative.

Available polling of independents and undecided/swing voters after Trump’s May 30, 2024 New York conviction shows a net negative reaction, not a sympathy-driven boost.

Key evidence:

• A Reuters/Ipsos poll summarized by Forbes found that among independents, 25% said the conviction made them less likely to vote for Trump, vs only 18% who said more likely—a clear net negative effect. Over half of independents in the same coverage said he should drop out of the race. (forbes.com)
• A Politico/Ipsos poll reported that 32% of independents were less likely to support Trump after the verdict, while only 12% were more likely; 53% of independents said he was guilty. (benzinga.com) Another Politico/Ipsos reading found 21% of independents less likely vs 5% more likely—again a net negative. (onnradio.com)
• An AP‑NORC survey found overall views of Trump and Biden were essentially unchanged from before the verdict, with independents divided on Trump and no sign of a broad sympathy surge; about half of U.S. adults approved of the conviction. (ny1.com)
• Morning Consult’s post‑verdict polling showed a majority of independents (about 52%) approved of the conviction and a similar share believed Trump had committed crimes, even as many opposed imprisoning him—suggesting limited sympathy for Trump himself, rather than for harsh punishment. (pro.morningconsult.com)
• A separate summary of polling noted that 52% of independents thought Trump should end his 2024 campaign after the conviction, underscoring admonishment rather than increased sympathy. (gvwire.com)

On Election Day, Trump did perform better with independents than in 2020 (e.g., Pew finds independents including leaners split 48–48 between Trump and Harris in 2024, after favoring Biden by 9 points in 2020; Edison data shows self‑identified independents 49% Harris, 46% Trump). (pewresearch.org) Other analyses suggest his share of the independent vote rose several points and that independents were pivotal in swing states. (thefulcrum.us) But these longer‑run shifts are driven by many factors (economy, immigration, Harris replacing Biden, etc.), and the direct polling on the conviction itself shows it initially hurt him with independents rather than generating net sympathy.

Because multiple independent polls consistently show that, among independents and undecided/swing voters, more people moved away from Trump than toward him in response to the New York conviction—and many wanted him to drop out—the prediction that the conviction would create a net sympathy effect and thus improve his standing with that group going into 2024 is best classified as wrong.

By late 2025, the tech/AI/SaaS space does not look like it entered a prolonged “slow contraction” starting in mid‑2024, nor like an initial AI mini‑bubble that burst and then deflated.

1. Valuations and sector size went up, not down.
Bessemer’s 2025 Cloud 100 report shows the leading private cloud/AI companies’ aggregate value jumping from about $820B in 2024 to about $1.1T in 2025, a ~36% increase, with valuations at all‑time highs and AI companies accounting for a rapidly growing share of that value. This is expansion, not contraction. (bvp.com)

2. Multiple compression was real, but ongoing since 2021, not a new mid‑2024 “reckoning.”
Cloud/AI revenue multiples have been drifting down from their 2021 peak for several years. Bessemer notes Cloud 100 revenue multiples falling from 26× (2023) to 23× (2024) to ~20× (2025), about 40% below the 2021 peak, while total valuations still hit records. (bvp.com) Public SaaS indices such as the BVP Nasdaq Emerging Cloud Index are trading around 7–8× revenue in 2024–25—roughly 65–70% below 2021 peaks, but described as having stabilized, not entering a new phase of contraction starting mid‑2024. (rockingweb.com.au) So multiples compressed, but that process began years earlier and coexisted with very strong growth in overall sector value.

3. Investor enthusiasm for AI remained extremely strong.
Far from a deflating mini‑bubble, AI names increasingly dominated markets through 2025. A widely cited overview of the “AI bubble” notes that in 2025 AI‑related enterprises were responsible for roughly 80% of U.S. stock‑market gains, that market concentration in a handful of mega‑cap tech/AI firms reached the highest levels in about 50 years, and that valuations were in the most stretched territory since the dot‑com era. (en.wikipedia.org) Parallel coverage of the AI trade describes “insatiable enthusiasm,” with Nvidia’s market cap briefly exceeding $4–5T, OpenAI’s private valuation rising into the hundreds of billions, and an IPO/financing boom for AI infrastructure and software firms. (financialcontent.com) Major platforms like Alphabet, Oracle, Palantir, and others saw huge 2025 gains explicitly tied to AI optimism. (reuters.com)

4. Net effect vs. the prediction.
The prediction was that, starting around mid‑2024, tech—especially AI and SaaS—was beginning a slow contraction driven by multiple compression and waning enthusiasm, marking the bursting of an initial AI mini‑bubble. What actually happened is better characterized as:

• Continued, gradual multiple normalization from 2021 extremes, but
• Record‑high aggregate valuations and sector size, and
• Intensifying AI hype and capital flows through 2025, not sustained cooling.

Taken together, the core thrust of the prediction—an AI/tech mini‑bubble starting to deflate into a slow contraction from mid‑2024 onward—has not played out by November 2025. The sector instead experienced an even larger AI boom with only intermittent corrections, so the forecast is best judged as wrong.

Why this prediction missed

1. No sustained U.S. economic slowdown from mid‑2024 onward. Real U.S. GDP grew at a 2.8% annualized rate in Q3 2024 and 2.3–2.4% in Q4 2024, with full‑year 2024 growth at 2.8%—above the post‑2000 average.(cnbc.com) After a brief 0.5–0.6% contraction in Q1 2025, GDP rebounded sharply to about 3.8% annualized in Q2 2025.(tradingeconomics.com) That pattern (solid 2H24 growth plus a one‑quarter dip followed by strong re‑acceleration) is not a “slow contraction from mid‑2024 onward.”

2. Enterprise / IT spending kept growing rather than falling. Gartner’s forecasts through 2024–2026 show rising worldwide IT outlays: up ~7.5% in 2024, with software and data‑center systems growing double‑digits, and further 7.9–9.3% growth projected for 2025, again led by software and AI‑driven infrastructure.(gartner.com) Gartner does note a 2025 “pause on net‑new spending” due to macro uncertainty, but explicitly says this is outweighed by ongoing AI digitization initiatives, not a broad pullback.(gartner.com) That is inconsistent with “reduced spending” and “reduced conversion of enterprise customers to buy anything.”

3. Tech stocks did not enter a sustained contraction; many hit record highs. The Nasdaq Composite was already near record levels in June 2024 and went on to set new all‑time highs in June 2025, alongside the S&P 500.(marketwatch.com) The Nasdaq‑100 likewise notched a record close of ~22,190 in June 2025 and later, by November 28, 2025, the Nasdaq was up ~21% year‑to‑date.(reuters.com) Sector ETFs focused on software and cloud—exactly the kind of enterprise tech he was talking about—show gains, not a sustained bear market: IGV returned about +27% in 2024 and is modestly positive in 2025 overall,(companiesmarketcap.com) while WCLD shows mid‑teens 1‑year returns as of September 2025 despite volatility.(myplaniq.com) Those trajectories contradict the idea of an enduring sector‑wide contraction.

4. Valuation multiples did not broadly compress across public tech. As of November 2025, IGV’s price/earnings ratio is around 52, a very elevated level by historical standards,(ishares.com) and market commentary through 2025 repeatedly highlights concerns about stretched AI‑driven tech valuations, not cheapened ones.(apnews.com) Even where there were pockets of de‑rating—e.g., a software ETF falling ~21% from a recent high and CIOs trimming budget growth expectations from 4% to 3% amid tariff worries—analysts still emphasized strong long‑term growth and only modest near‑term caution.(barrons.com) That is a far cry from broad, lasting multiple compression across tech.

Taken together, the data show: (a) continued macro growth with just a brief 2025 soft patch, (b) rising, not falling, enterprise tech spending, and (c) technology equity markets that, while volatile, reached record levels and retained high valuations rather than entering a long, contractionary, multiple‑compression phase. That makes the prediction wrong in both its macro and sector‑specific components.

Evidence since the May 2024 superalignment departures matches the core of Friedberg’s prediction: former OpenAI safety/superalignment staff have been formally questioned by government oversight bodies, and that has produced new public information about OpenAI’s frontier models and safety processes.

Key points:

1. Former superalignment member testified before the U.S. Senate.
   In September 2024, the U.S. Senate Judiciary Subcommittee on Privacy, Technology, and the Law held a hearing at which former OpenAI employees testified about AGI risks. One of them, William Saunders, is a former OpenAI researcher who spent three years on the superalignment team before resigning in early 2024 over safety concerns.(techtarget.com) This matches the scenario Friedberg described—government overseers bringing in members of “this team” to ask detailed questions.

2. His testimony revealed previously non‑public internal safety and model details.
   In that Senate hearing, Saunders described:

   • That AGI might be feasible in as little as three years, explicitly tying this to OpenAI’s recent frontier model (GPT‑o1).
   • That OpenAI had not prioritized internal security, and there were extended periods where vulnerabilities allowed employees to bypass access controls and steal its most advanced models, including GPT‑4.
   • That OpenAI had “repeatedly prioritized speed of deployment over rigor,” leaving serious safety gaps.(techtarget.com)
     These details about internal access‑control weaknesses, theft risk for GPT‑4–level systems, and the way safety reviews were traded off against deployment were not part of OpenAI’s prior public messaging; they emerged through sworn testimony in an official oversight setting.

3. Regulatory / oversight pressure also surfaced internal policies that discouraged safety whistleblowing.
   In July 2024, lawyers for a group of OpenAI whistleblowers sent a detailed letter and formal complaint to the U.S. Securities and Exchange Commission (SEC) alleging that OpenAI’s employment, severance, and NDA practices illegally restricted employees from warning regulators about grave safety risks. The letter describes specific contract clauses (waiving federal whistleblower compensation, requiring company permission before talking to regulators, threats of “severe sanctions”) and notes that the SEC has responded to the complaint.(washingtonpost.com) That complaint and the surrounding coverage publicly exposed internal legal and safety‑governance practices that had not been disclosed before—and they exist precisely in the context of regulator oversight.

4. All of this unfolded soon after the May 2024 superalignment resignations, within the 1–2 year window.
   Ilya Sutskever and Jan Leike announced their departures and the effective disbanding of OpenAI’s superalignment team in mid‑May 2024.(ndtv.com) The SEC whistleblower complaint (July 2024) and the Senate testimony by Saunders (September 2024) both occurred within months of those departures and well inside Friedberg’s 1–2 year horizon.

5. Nuance: leaders vs. team members.
   So far, there’s no public record that regulators have formally interviewed Sutskever or Leike themselves in hearings or regulatory reports. The clearest on‑the‑record interactions involve other members of the superalignment/safety ecosystem (e.g., Saunders, plus ex‑OpenAI governance staff like Daniel Kokotajlo in related advocacy and testimony).(openaifiles.org) Still, the spirit of Friedberg’s prediction was that once “this team” left, regulators/oversight would pull them in, ask hard questions about the state of the tech and safety, and “some things are going to start to come out.” That’s exactly what happened with Saunders’ Senate testimony and the SEC‑focused whistleblower actions.

Because government oversight bodies have indeed brought in former OpenAI superalignment/safety staff, and those proceedings have surfaced new, previously non‑public information about OpenAI’s cutting‑edge models and safety/security processes, within the predicted timeframe, the prediction is best classified as right (even if the most visible testimony so far has come from team members rather than the two most senior leaders).

Evidence since mid‑2024 shows the beginnings of exactly the trend Friedberg described, both on the demand side (health and chemical concerns about plastics in food) and the supply side (food‑related businesses rolling out low‑/no‑plastic and bioplastic options and marketing them as safer or premium).

1. Rising concern about plastic chemicals in food and explicit focus on phthalates
   Consumer Reports’ widely covered testing found phthalates and bisphenols in nearly all sampled U.S. supermarket foods and fast foods, often at high levels, and highlighted associated health risks, sparking substantial media coverage and consumer guidance on reducing exposure. (food-safety.com) In early 2025, Minnesota introduced bill SF 188 to require packaged‑food manufacturers to test for phthalates and report results, explicitly framing phthalates in food contact materials as a serious health issue. (foodandwine.com) Environmental and public‑health groups also sued the U.S. FDA in 2025 over its refusal to restrict phthalates in plastic food packaging, again centering the chemicals’ reproductive and developmental risks. (theguardian.com) Parallel articles and consultancy pieces describe phthalates in plastic packaging as a major public‑health risk, reinforcing both regulatory and consumer pressure. (condor-consultancy.com) Separately, industry analysis notes a strong shift toward plasticizer‑free and phthalate‑free food wraps, citing surveys where U.S. consumers explicitly seek “non‑toxic” and phthalate‑free food wrap and forecasts of double‑digit growth for plasticizer‑free films through 2025. (ecoartfullife.com) This is exactly the kind of rising attention to “low phthalate” food contact that Friedberg anticipated.

2. Food and packaging companies actively marketing “low/no plastic” and plastic‑free solutions as premium or value‑add
   Major retailers and brands have introduced packaging changes for food products that are explicitly marketed as plastic‑free or low‑plastic. Examples include: Aldi’s Everyday Essentials porridge oats in the UK, relaunched in 2024 with 100% paper, plastic‑free packaging and promoted as part of the chain’s move away from plastic; (trendhunter.com) Sainsbury’s shift to pulp‑paper trays for salmon and trout that cut plastic by ~70%, positioned as reducing reliance on plastic and improving recyclability; (cleanthesky.com) and Stora Enso’s Trayforma BarrPeel board for vacuum‑packed meats and cheese, specifically described as fresh‑food packaging with low plastic content (about 90% fiber) to reduce fossil‑based materials and meet consumer demand for sustainable options. (packaging-journal.de) Packaging suppliers now routinely promote “low‑plastic solutions” for food (e.g., Pyroll’s paper laminates marketed as natural, responsible, and explicitly labeled ‘low‑plastic solutions – smaller carbon footprint’). (pyroll.com) India‑based FMCG major ITC has adopted a three‑pillar “No plastic, Better Plastic, Less Plastic” framework and launched flour and biscuit packs with 100% paper outers and bio‑based coatings, emphasizing biodegradable, fiber‑based packaging for food as an eco‑friendly, premium alternative to conventional plastic laminates. (packagingsouthasia.com) These are concrete examples of food‑industry players “showing up with solutions” and using low‑/no‑plastic packaging as a selling point.

3. Bioplastic and alternative materials in food‑related applications are clearly in a commercial ‘wave’ phase
   Multiple independent market reports show that bioplastic and sustainable plastic packaging are now a fast‑growing, multi‑billion‑dollar segment, with food and beverages as the dominant end use. A global sustainable plastic packaging report values that market at roughly $86.6B in 2024, projecting $122.4B by 2029, explicitly noting widespread use of biodegradable materials like PLA in food containers and flexible packaging. (businesswire.com) Dedicated bioplastic‑packaging studies estimate 2024–2025 market sizes on the order of $10–20B with CAGRs in the mid‑teens to >20%, and consistently report that food and beverages are the leading or fastest‑growing application segment. (grandviewresearch.com) Packaging‑industry coverage notes that a large share of global bioplastics volume (around 40–45%) is used in packaging, much of it for food, as brands react to environmental and regulatory pressures. (knowledge-sourcing.com) On the product side, there are numerous concrete launches: biodegradable PHA‑ and wood‑fiber strawberry punnets intended to replace hundreds of millions of plastic clamshells; (theaustralian.com.au) Papacks’ fully fiber, plastic‑free bottles being commercialized for beverages in North America; (foodnavigator-usa.com) and extensive adoption of compostable cutlery and serviceware in quick‑service chains, with case studies showing large chains replacing plastic utensils with compostable alternatives and seeing higher customer satisfaction. (digitaljournal.com) An industry article on compostable cutlery cites a 2025 Technomic study finding that 81% of consumers are more likely to repurchase from brands using compostable packaging and 62% will pay a premium for visibly eco‑conscious options, confirming that these materials are being positioned as higher‑value, health‑ and environment‑conscious choices rather than commodity plastics. (digitaljournal.com)

4. Explicit marketing around health / “safer than traditional plastics”
   Some food‑related packaging offerings now explicitly emphasize the absence of problematic chemicals as a differentiator, not just environmental benefits. Asia Pulp & Paper’s Foopak Bio Natura food‑service line is marketed in the U.S. as plastic‑free, compostable and recyclable and “free from harmful chemicals found in traditional plastic packaging,” in response to survey data showing 83% of Americans increasingly worried about single‑use plastics’ harms to ecosystems and human health, with 68% willing to pay more for sustainable products. (packworld.com) Consumer‑facing guidance and product round‑ups increasingly highlight plastic‑free, BPA‑free and phthalate‑free food storage and lunch containers as “non‑toxic” or “healthier” options, reflecting and reinforcing the health framing Friedberg anticipated, even if much of this is in food‑adjacent products (wraps, containers, lunch boxes) rather than the food itself. (nakedpantry.co)

5. Magnitude and caveats
   It is still true that bioplastics remain a small fraction of overall plastic volume (on the order of a few million tonnes vs. hundreds of millions of tonnes of fossil‑based plastics worldwide), and most mainstream packaged foods do not yet carry explicit front‑of‑pack labels such as “low phthalate” in the way that “organic” or “non‑GMO” labels exist. (ft.com) But Friedberg’s prediction was about the start of a wave in the mid‑2020s: food companies paying attention to plastic/chemical concerns, positioning low‑/no‑plastic or bioplastic options as premium responses, and a broader commercial push of bioplastic alternatives in food‑related applications. On those terms, the evidence by late 2025—regulatory and media focus on phthalates, measurable consumer willingness to pay more for safer/sustainable packaging, rapidly growing bioplastic packaging markets led by food and beverage, and many concrete low‑/no‑plastic and bioplastic packaging launches framed as healthier or more responsible—supports that this wave has indeed begun.

Given that the timeframe he mentioned (“starting in the mid‑2020s”) is exactly where we are, and the observable industry and market trends match his described pattern, the prediction is best classified as right, acknowledging that the shift is underway rather than complete.

As of November 30, 2025, there is no evidence that Ohalo’s Boosted Breeding technology has been applied across nearly every major crop worldwide or that it has had a visible, system‑wide impact on global food prices and sustainability metrics—but the time horizon for such a sweeping prediction is inherently long, so it is too early to call it definitively right or wrong.

Key points:

1. Current scope of deployment is narrow, not “nearly every major crop worldwide.”

   • Ohalo’s own materials describe Boosted Breeding as a platform that can be applied to nearly every crop on Earth, but they highlight active programs mainly in potatoes, strawberries, and some other specialty or pilot crops, not a broad, global rollout across all major cereals and oilseeds.(ohalo.com)
   • The May 23, 2024 press release and subsequent coverage emphasize that the technology has been discovered and that Ohalo is implementing Boosted Breeding to drive breakthroughs across a “multitude of crops” with partners and plans to expand, which indicates early rollout and R&D rather than mature, global adoption.(prnewswire.com)
   • Concrete commercial efforts mentioned publicly include Boosted Potato (true potato seed and value‑added traits) and the Ohalo Strawberry Consortium launched in 2025, again pointing to specific crops and regions rather than most major crops worldwide.(ohalo.com)

2. No evidence that the technology has materially lowered global food prices.

   • Global food prices over 2024–2025 have been driven mainly by macro factors like weather, energy, and geopolitical shocks. The FAO Food Price Index in 2025 is volatile and in many months higher year‑on‑year than in 2024, even if well below the 2022 peak.(tradingeconomics.com)
   • There is no attribution in FAO, Reuters, or other major analyses of food prices to Ohalo’s technology, which is still at an early‑adoption stage; any acreage under Boosted varieties is far too small (and geographically limited) to measurably affect global commodity prices.

3. Sustainability and yield claims remain mostly at the trial / promise stage.

   • Ohalo reports large yield gains (50–100%+ in early trials) and theoretical sustainability benefits (less land, water, and inputs per unit of food), but these are framed as potential or early‑trial outcomes, not as measured, global‑scale impacts across major crops.(ohalo.com)
   • Job postings and materials emphasize field trials, early adopter farmers, and development of agronomic protocols for true seed systems, again indicating that the technology is in scale‑up and validation rather than mature, industry‑wide deployment.(jobs.valorcapitalgroup.com)

4. Why this is “inconclusive” rather than “wrong.”

   • The prediction did not specify a time frame. Transforming breeding systems and seed supply chains across nearly every major crop worldwide would, in realistic agricultural timelines, likely take well over a decade. Only ~18 months have passed since the podcast (May 17, 2024), which is far too short to judge such a structural, global claim.
   • What we can say confidently is that the predicted end state has not yet materialized by late 2025; what we cannot say is that it will not materialize over the longer term. Hence, the correct classification is that it is too early to evaluate definitively.

So, the prediction is unfulfilled so far, but given the inherently long adoption cycle in global agriculture and the lack of a stated deadline, it is best categorized as inconclusive (too early) rather than clearly right or wrong.

As of November 30, 2025, there is not enough independent, quantitative evidence to say whether Ohalo’s boosted‑breeding crops actually use less water, land, and energy per unit of output in real‑world farming.

What we do know:

• Ohalo and David Friedberg report very large yield gains (often 50–100%+ in early potato trials) from Boosted Breeding, which would mechanically reduce land required per unit of output if input use per acre stayed similar. These claims are based on company trials and press materials, not yet large-scale independent field data. (ohalo.com)
• Friedberg has explicitly claimed that the technology will enable “more food per acre using less water, less land, less fertilizer per unit output,” but this is framed as the potential impact of the platform, not as results from long‑term commercial deployment. (freethink.com)
• Ohalo is still early in commercialization: potatoes and strawberries are in programs and consortia, and an almond variety (FruitionOne) with a projected ~50% reduction in water per almond is only in trial orchards, with first commercial deliveries targeted for 2027. (ohalo.com)
• Regulatory milestones (e.g., USDA APHIS review of an Ohalo potato line) and initial product offerings show the tech is moving toward market, but they do not provide measured data on water, energy, or land use per ton of output relative to conventional varieties. (agtechnavigator.com)

Because the public record so far is dominated by company‑reported trials and forward‑looking statements, and there are no peer‑reviewed or large‑scale independent studies quantifying water, land, and energy use per unit yield for Ohalo’s boosted crops versus conventional varieties, Friedberg’s multi‑part efficiency prediction cannot yet be judged as clearly correct or incorrect. It remains too early to call.

As of November 30, 2025, there is not enough real‑world data to determine whether Ohalo’s Boosted Breeding crops actually use significantly less fertilizer per pound of harvested output in commercial settings.

Key points:

1. Stage of deployment: After unveiling Boosted Breeding in May 2024, Ohalo reported large yield gains (50–100%+) in early internal trials, but described the platform as moving toward commercial trials and looking for partners, not yet in broad commercial deployment.(agtechnavigator.com) A June 2024 Freethink piece explicitly says Ohalo is preparing for commercial trials and seeking partners.(freethink.com) A 2025 job posting for a "Field Technology Development & Agronomy Lead" to support "early adopter farmers" with true potato seed further indicates they are still building out field systems and protocols rather than operating at mature commercial scale.(s32.getro.com)

2. Claims vs. measured fertilizer data: Friedberg and Ohalo repeatedly claim that Boosted Breeding will enable more food per acre with less fertilizer per unit output, echoing the prediction.(freethink.com) However, public materials (press releases, interviews, product pages) focus on yield, disease resistance, true seed, and reduced pesticide use; they do not provide quantified fertilizer‑input data (e.g., N/P/K rates or fertilizer per ton of yield) from commercial or large‑scale field trials.(prnewswire.com) No independent agronomic studies or multi‑season commercial trial results reporting fertilizer use per unit yield for Boosted crops appear in the literature or trade press.

3. Timeline for full commercial products: The only clearly dated commercial roadmap item with regulatory detail, the FruitionOne self‑fertile Nonpareil almond variety, has trial orchards underway with first commercial deliveries not expected until 2027, again suggesting the broader product portfolio is still pre‑mass‑deployment.(prnewswire.com)

Because:

• Boosted Breeding crops are only in early adopter / preparatory commercial‑trial phases, and
• There is no published evidence yet on actual fertilizer applied per pound of harvested crop in those trials or deployments,

we cannot say whether fertilizer intensity has in fact fallen "significantly" as predicted. The prediction is therefore too early to evaluate, not clearly right or wrong at this time.

The prediction is framed on a multi-decade horizon ("a new era" with "pretty profound" impacts on drug development and human health), but we are only about 1.5 years past the May 2024 release of AlphaFold 3, far too early to evaluate long‑run, large‑scale health and industry impacts.

Evidence so far is directionally supportive but preliminary:

• AlphaFold‑based methods are being actively refined for drug design; for example, the AF2RAVE approach explicitly combines AlphaFold with physics-based simulations to improve drug-relevant protein conformations, with researchers describing this as enabling faster drug discovery and potentially personalized medicine, but they themselves characterize this as an early-stage step toward that vision. (cmns.umd.edu)
• AI-driven drug discovery companies (including those building on AlphaFold-style models) have AI‑designed drug candidates entering or approaching human clinical trials, which is a major milestone but still only the beginning of the clinical pipeline. (aitechsuite.com)
• Market analyses project rapid growth of AI in drug discovery through 2030 and describe it as transformational for timelines and costs, indicating strong industry uptake but not yet demonstrating realized, population-scale health outcomes. (stocks.observer-reporter.com)

However, demonstrating “profound, large-scale impacts on human health and the biopharma industry” requires:

• Successful completion of Phase II/III trials for multiple AI-designed drugs
• Regulatory approvals and widespread clinical adoption
• Clear, measurable improvements in morbidity, mortality, or cost at scale across many indications

Those outcomes typically take 10–15 years or more from early discovery to broad clinical impact, and as of late 2025 the field is still in the early validation phase. Current data show promise but cannot yet confirm whether the long-run, transformative impact Friedberg predicted will actually materialize.

Because the forecast explicitly concerns the next couple of decades and we are only a short way into that period, with crucial clinical and economic results still pending, the prediction cannot be judged as right or wrong at this time.

As of November 30, 2025, this prediction is about a long‑term future capability and clinical impact, and there is not yet enough time or evidence to say it is clearly right or clearly wrong.

What has happened so far

1. AlphaFold 3 and related models can model complex interactions, but mainly in silico and preclinical contexts.

   • DeepMind/Isomorphic Labs’ AlphaFold 3 (2024) extends structure prediction to protein–protein, protein–DNA/RNA, and protein–ligand complexes, improving the ability to model binding interactions across chemical space.
   • However, these are prediction tools, not validated end‑to‑end drug‑discovery pipelines that guarantee safe, specific reprogramming of human cells.

2. In‑silico search over chemical and biological space is real and rapidly improving.

   • Many groups and companies now use generative and predictive models (including structure predictors like AlphaFold, diffusion models, and reinforcement learning) to design small molecules, biologics, and RNA constructs. Some are applied to aging-related pathways or senolytics in preclinical work.
   • These tools support exploration of candidates but do not yet reliably deliver clinically proven longevity or rejuvenation therapies.

3. Yamanaka‑factor / partial reprogramming for rejuvenation is still preclinical.

   • There is active research using OSKM and related factors for cellular rejuvenation and partial reprogramming in mice and cell culture, and several startups (e.g., Altos Labs, NewLimit, etc.) are pursuing this area.
   • As of 2025, there is no approved human therapy that safely “de‑ages all the cells in the body” using Yamanaka‑style reprogramming, and no widely accepted clinical protocol derived from an AlphaFold‑like in‑silico combinatorial search.

Why the prediction is “inconclusive” rather than wrong

• The claim is essentially that in the future, systems like AlphaFold 3 will let us computationally search over combinations of molecules that can safely reprogram cells and unlock powerful rejuvenation/longevity therapies.
• This is a directional, long‑horizon forecast about what will be possible, not about something that should already have occurred by 2025.
• Current evidence shows early enabling steps (powerful structure prediction, AI‑assisted design, active partial‑reprogramming research), but not the end state (validated, safe, clinically deployed rejuvenation therapies discovered this way).

Because the timeframe is open‑ended and the core claim concerns capabilities and therapies that plausibly lie decades away, it is too early to judge it as right or wrong. Hence the status is “inconclusive (too early)”.

The “fake Chamath” handle referred to in the episode is almost certainly @ChamathWarriors – “Chamath Palihapitiya’s burner (parody)” on X/Twitter. The episode transcript matches this: Friedberg jokes about Dave Sacks “being fake Chamath” and says, “My guess is that Twitter handle is about to get popular again.”(podscripts.co)

What we can see empirically:

• The @ChamathWarriors account has been active and widely-cited well before the May 3, 2024 episode. Its memes were quoted in media pieces and blogs in 2022–2023 (e.g., on ChatGPT and Pixelmon) and in political/tech commentary, showing it already had notable reach.(moguldom.com)
• As of a late‑2025 snapshot, @ChamathWarriors shows ~7,000 tweets and ~16,000 followers, indicating a moderately popular, consistently active parody account, not an abandoned or tiny one.(twstalker.com)
• After the May 2024 episode, the account continues to be referenced in third‑party content: for example, an investing newsletter highlights its memes in October 2024 (“Perfect last‑minute costume idea”) and again in February 2025 (“Dave said go all‑in”), suggesting it’s still part of the broader finance/tech meme ecosystem and now explicitly tied into All‑In in at least one meme.(becomeabetterinvestor.net)

However, the core of the prediction is about a “significant increase in popularity and activity” after the episode. To judge that, we’d need things like:

• follower counts or engagement metrics before May 3, 2024 vs. after;
• or reporting/commentary that clearly notes a resurgence (“this account suddenly blew up after being outed on the All‑In podcast”).

Publicly accessible sources don’t provide historical follower data for @ChamathWarriors, and no coverage explicitly documents a distinct popularity spike or revival around that time. All the evidence shows is steady, ongoing popularity from 2022 through 2025, with mentions both before and after the episode, not a clearly identifiable “re‑becoming popular” moment.

Because:

• the account was already popular enough to be quoted in mainstream and niche outlets before the prediction; and
• we lack quantitative or clear qualitative evidence of a step‑change in popularity or activity afterwards;

it’s not possible to definitively say whether Friedberg’s “about to get popular again” call was correct or incorrect. The account remained active and visible, but whether there was a significant renewed surge is not verifiable from available data.

So the fairest classification is ambiguous: enough time has passed, but the necessary metrics to confirm or refute a “significant increase” don’t appear to be publicly documented.

As of November 30, 2025, the forced sale of TikTok’s U.S. business has not actually been completed, only negotiated in principle.

After Congress passed the Protecting Americans from Foreign Adversary Controlled Applications Act in April 2024, ByteDance faced a divest-or-ban deadline, upheld by the Supreme Court in TikTok, Inc. v. Garland. TikTok temporarily went offline in January 2025, after which President Trump delayed enforcement and began negotiating a divestiture framework. 【0search13】【3search19】

By mid‑September 2025, reporting described only a “framework of a deal” for the U.S. operations of TikTok to be sold to a U.S.‑led investor consortium (Oracle, Silver Lake, Andreessen Horowitz, others), with Trump extending the ban/divestment deadline to December 16, 2025. These accounts consistently characterize the arrangement as pending or to be finalized, not closed. 【1news11】【1news13】【3search18】【0news16】【0news17】

Because the divestiture/ownership transfer has not yet closed, the condition in Friedberg’s prediction ("if it does go through" – i.e., a completed forced sale/auction of TikTok’s U.S. business) has not been met. Without a completed deal, there is naturally no post‑closing media cycle yet that would focus on (1) disclosed advisory fees to investment banks for running a completed auction, and (2) retrospective analyses on how the tech and media landscape has been reshaped by a finalized TikTok ownership change.

Some coverage, including a WSJ‑based exclusive cited by Voz.us, discusses a proposed multi‑billion‑dollar fee the U.S. government might charge buyers for brokering the TikTok deal, and only mentions typical investment‑bank fee percentages as contextual background; it is not a WSJ feature about banks’ auction fees on a consummated transaction. 【2view0】

Given that the triggering event (a completed forced sale of TikTok U.S.) has not yet occurred, the prediction cannot be definitively judged right or wrong at this time; it remains contingent on future developments.

Friedberg’s on-air statement was:

“we had within 72 hours, I think, more applications than we have seats… And in the next week, we’ll start to respond to people… it’s going to be done in order of when it’s received. And they’re going to start processing applications this week.” (podscripts.co)

Key pieces to verify:

1. Timing – “in the next week we’ll start to respond / start processing applications this week.”
   • The episode is dated April 19, 2024, so this implies initial responses and processing should begin roughly by April 26, 2024. (podscripts.co)
   • A Reddit thread from May 9, 2024 shows at least one applicant saying they had not heard back yet, but another commenter says they have seen people on X (Twitter) who “have been accepted and paid the admission fee,” indicating some acceptances were going out by early May. (reddit.com)
   • Public web data (tweets, posts, press) does not clearly show when the very first acceptance emails went out, only that some people were accepted by May 9. That’s consistent with the prediction but doesn’t prove responses actually started within the specific one‑week window after April 19.

2. Oversubscription claim – “within 72 hours… more applications than we have seats.”
   • There is no independent public reporting that confirms or refutes the exact 72‑hour oversubscription metric. Later coverage simply notes the All‑In Summit is an invite‑only, sold‑out conference with around 1,800–1,950 attendees paying about $7,500 each in 2024, which is compatible with strong demand but not specific enough to verify the 72‑hour detail. (en.wikipedia.org)

3. Processing “in order of when it’s received.”
   • The application and review process is run privately by the organizers; available anecdotes (e.g., some applicants waiting while others are already admitted) don’t establish whether they actually followed strict first‑come‑first‑served ordering or made exceptions (e.g., sponsors, VIPs, scholarships). (reddit.com)

Because:

• The relevant actions (exact start date of responses and internal processing order) happen entirely inside a private system; and
• Public traces only show that some acceptances existed by May 9, 2024, without pinning down when processing began or how strictly ordering was followed,

there isn’t enough verifiable external evidence to say the prediction was clearly correct or clearly wrong. It can’t be determined even though the event is long past, so the appropriate classification is “ambiguous.”

The transcript of the April 19, 2024 episode confirms Friedberg saying about All‑In Summit speakers: “In a week, we’ll announce a bunch together,” during the summit update segment. (podscripts.co) Earlier, the April 12 episode described All‑In Summit 2024 and explicitly noted that a lineup of notable speakers was still “to be announced,” so a first speaker list had not yet been made public by then. (allin.onpodcastai.com) However, in searching news coverage, blogs, the All‑In / All‑In Summit web pages, and later episode summaries and transcripts, I can find no dated record of a public “batch” announcement of confirmed 2024 Summit speakers around a week later (roughly April 26, 2024). (en.wikipedia.org) Those sources do confirm the summit, its dates, and a long list of ultimate speakers, but none specify when the initial speaker cohort was publicly announced. (en.wikipedia.org) Because such an announcement could have been made via channels that are not well indexed (e.g., email blasts or certain social posts) and I cannot find clear evidence either that it happened on the stated timeline or that it did not, the accuracy of the prediction cannot be determined from available public records.

The prediction matches what actually happened.

1. The fix did stall past April 15, 2024. A $78B bipartisan tax bill (including restoration of immediate R&D expensing under Section 174) passed the House in January 2024 but remained stuck in the Senate as Americans filed by the April 15 deadline, and senators themselves warned it might be “too late” to salvage it for that filing season. (wsj.com) Later reporting shows the Senate effectively killed this bill, so no broad retroactive R&D fix for 2022–2023 passed in that Congress. (wsj.com) During 2023–2024 filings, businesses therefore had to capitalize and amortize R&D over five or fifteen years instead of expensing it immediately.

2. Many small, R&D‑intensive firms did experience significant financial strain from this rule (i.e., taxes on “phantom” profits).

   • A 2025 survey of founders by startup bank Mercury found that 73% said the R&D amortization change had a negative financial impact, and 44% reported operational changes such as cutting R&D budgets, slowing or reducing hiring, delaying/canceling projects, offshoring, and even layoffs. (mercury.com) These respondents are predominantly tech startups.
   • The U.S. Chamber of Commerce reports that, since mandatory R&D amortization began in 2022, many small and midsize businesses have suffered cash‑flow and liquidity problems and been forced to take high‑interest loans, raise prices, and stop hiring just to pay their higher tax bills. (uschamber.com)
   • A Senate small‑business roundtable summarized survey data showing a median 32% increase in tax bills (about $59,000) for small firms subject to the new R&D rules; 35% had to borrow to pay the tax, and 19% said their firm might go out of business, with additional impacts like reduced investment, hiring cuts, and layoffs. (congress.gov)
   • Industry letters (e.g., from manufacturing and startup groups) describe the R&D amortization requirement as causing “significant cash flow impacts” on startups and small businesses, forcing them to delay investments, forego hiring, and take out loans, and warn that “tens of thousands of jobs are at risk” if it remains. (nffs.org) The National Defense Industrial Association specifically notes that losing most of the R&D deduction creates dramatically higher tax bills that many companies, especially small defense‑related businesses, will struggle to manage, tying it to national‑security supply chains. (nationaldefensemagazine.org)
   • Policy and news analyses on Section 174 explicitly explain that amortization can make unprofitable startups appear profitable for tax purposes, resulting in tax bills despite having no real earnings — the textbook definition of “phantom” taxable income. (axios.com)

3. Timing relative to the 2023–2024 tax years. The amortization requirement has applied since tax years beginning in 2022, so its first big impact hit 2022 and 2023 returns (filed in 2023 and 2024), with continuing effects into 2024 filings, precisely the window the prediction referred to. (cbh.com) A later 2025 reform (creating new Section 174A and allowing some retroactive small‑business relief via amended returns) only arrived after this period and does not negate the substantial cash strain and operational damage many small tech, life‑sciences, and defense‑related firms reported during 2023–2024.

Given (a) the confirmed Senate stall beyond April 15, 2024, and (b) strong empirical evidence that a large number of small, R&D‑intensive U.S. businesses faced large tax increases, cash‑flow problems, borrowing needs, and cuts to R&D and hiring because of Section 174 amortization, the prediction that this stall would “hurt a lot of small businesses” through taxes on phantom profits is best judged right.

Legal/policy part of the prediction

• The January 2024 bipartisan framework and H.R. 7024 (“Tax Relief for American Families and Workers Act of 2024”) would have restored immediate expensing only for domestic R&D while keeping foreign research expenditures subject to 15‑year amortization under §174. House and practitioner summaries explicitly state that foreign R&D costs would “continue” to be amortized over 15 years. (congress.gov)
• That specific 2024 bill ultimately stalled, but in 2025 the One Big Beautiful Bill Act (OBBBA) was enacted. OBBBA rewrote §174 so that it now applies only to foreign research, which “continues to be amortized” over 15 years for tax years beginning after December 31, 2024. A new §174A restores full expensing for domestic research. IRS guidance and technical summaries confirm this structure. (irs.gov)
• Multiple tax advisories after OBBBA describe current law as: (1) permanent or effectively permanent full expensing for domestic R&D, and (2) no change for foreign R&D, which must still be capitalized and amortized over 15 years, with no scheduled sunset. (manercpa.com)

This matches Friedberg’s core claim that the legislative fix would (and now does) leave in place a rule forcing US companies to spread offshore R&D over 15 years while giving much more favorable treatment to domestic R&D.

Economic consequences part of the prediction

• Practitioners and IRS commentary note that the §174 capitalization rules increased taxable income and reduced cash flow for R&D‑intensive firms when they first took effect, precisely because immediate deductions were replaced with slow amortization. (irs.gov)

• Under current law, a US company that does a growing amount of offshore R&D must still capitalize and deduct those costs slowly over 15 years, while domestic costs can be expensed. This generally means taxable income is higher than it would be under immediate expensing and can exceed “economic” profit, especially in growth phases when R&D spending is ramping up. Analysts explicitly warn that the foreign‑only capitalization regime may drive firms to relocate R&D to the US for tax reasons, which reflects the same underlying distortion Friedberg described. (grantthornton.com)

• His stronger language (“any” such business will “always” pay tax even at breakeven) is somewhat hyperbolic, because net operating losses, credits, or other deductions can offset tax in particular cases. But the directional mechanism—that a 15‑year amortization rule on foreign R&D raises taxable income relative to economic profit for growing firms with significant offshore R&D—is correct and is recognized in the professional commentary.

Given that (1) the law has in fact evolved so foreign R&D must be amortized over 15 years indefinitely while domestic R&D is expensed, and (2) this structure does exactly what he said in terms of tax distortions (even if not literally in every single case), the prediction is best classified as right overall.

The prediction covers a long window — “the decade ahead” from roughly 2024 through 2034. As of now (November 30, 2025), only about a year and a half of that period has elapsed, so it’s too early to judge whether the full claim (sustained, decade‑long pressure leading to increasingly innovation‑distorting business taxes) is right or wrong.

Some early context:

• U.S. federal debt and projected deficits remain high, and official budget projections from bodies like the Congressional Budget Office anticipate continued upward pressure on spending, especially from entitlements and interest costs, which in turn imply pressure for higher revenues over the long term. This is directionally consistent with the motivation behind the prediction, but doesn’t by itself confirm the specific outcome about how new taxes will target innovative businesses.
• There have been ongoing debates and proposals around corporate minimum taxes, digital/service taxes, and higher taxes on large or highly profitable firms in tech and other sectors, but these are incremental and politically contested. It’s not yet clear that we are in a sustained regime where “more and more tax and all these weird ways that can really hurt industry” is definitively the pattern for the remainder of the decade.

Because the claim is (1) medium‑term macro‑fiscal, (2) explicitly about trends over an entire decade, and (3) stated in qualitative terms about how much new tax pressure will hurt innovation‑driven industries, we don’t have enough elapsed time or clear evidence to decisively deem it right or wrong. The appropriate classification as of 2025 is therefore “inconclusive (too early).”

By late 2025, text‑to‑music systems like Suno and Udio have become widely used by both creators and consumers: Udio reports millions of users since its 2024 launch, its tracks have gone viral and even entered national charts (e.g., an Udio‑generated song reaching the German Top 50), and Suno‑based songs have driven major TikTok/streaming controversies and licensing deals with the three major labels. (cirrkus.com) AI‑generated artists and songs are also charting commercially (for example, an AI gospel artist topping Christian and Gospel digital charts and the viral, fully AI‑generated song “We Are Charlie Kirk” topping Spotify’s viral songs chart), indicating exactly the kind of prolific use by musicians and ordinary users Friedberg described. (nypost.com) On the legislative side, Tennessee enacted the ELVIS Act in 2024, explicitly framed as the first U.S. law to protect musicians from AI voice cloning and audio deepfakes, directly targeting generative music/voice tools. (en.wikipedia.org) In Congress, lawmakers introduced the No AI FRAUD Act to create a federal right over one’s voice and likeness against AI fakes—promoted by its sponsors and the RIAA as necessary to protect music artists from AI‑driven impersonation—and additional bills like the TRAIN Act seek transparency around unauthorized AI training on creators’ works. (salazar.house.gov) Abroad, the EU’s AI Act and national initiatives such as Denmark’s proposed law granting people rights over AI‑generated imitations of their body, face and voice, plus UK parliamentary committees explicitly focusing on AI’s use of copyrighted music (accompanied by a silent protest album by 1,000 artists), show legislators actively grappling with generative‑AI impacts on music and related rights. (en.wikipedia.org) Taken together, rapid mainstream adoption of generative music tools and the emergence of music‑focused or music‑driven AI laws and bills match Friedberg’s forecast that widespread use would force legislators to confront and debate regulation of generative music AI, so the prediction is essentially correct by late 2025.

Evidence strongly supports Friedberg’s prediction that the spike in cocoa prices would lead major chocolate makers to shrink products and raise effective prices, with consumers seeing smaller and more expensive chocolate.

• Cocoa price spike: Cocoa futures roughly tripled in 2024, at points hitting around $10,000 per metric ton—far above prior norms—creating intense cost pressure on chocolate manufacturers.(benzinga.com)
• Early concrete shrinkflation tied to cocoa: Reporting in March 2024 noted that Mars had already reduced the size of its standard Galaxy bar by 10 grams while keeping the price the same, explicitly in response to soaring cocoa costs.(dairynews.today) Similar moves were later documented for Mondelez’s Cadbury Freddo multipacks, which were cut from five to four bars while the price stayed at £1.40, again justified by sharply higher cocoa and other input costs.(thescottishsun.co.uk)
• Size cuts and price hikes across big manufacturers: By late 2024–2025, major brands including Nestlé and Mondelez were reported to have reduced product sizes, raised prices, and in some cases lowered cocoa content in response to the cocoa crisis, with Christmas and everyday chocolate described as noticeably more expensive despite smaller packs.(thetimes.com) Analyses of the confectionery sector similarly describe strategies of shrinkflation (smaller packs at same price) and skimpflation (less cocoa per product) as industry‑wide responses to high cocoa costs.(foodprocessing.com.au)
• Consumers seeing smaller, costlier chocolate on shelves within months: Halloween 2024 coverage in Canada and North America notes that fun‑size chocolate bars and Halloween candy boxes contained fewer or smaller pieces than before, with cost per item up 5–10% versus the prior year, explicitly framed as shrinkflation amid sharply higher cocoa prices.(retail-insider.com) A Wells Fargo/industry analysis on the 2024 Halloween season likewise highlighted that record cocoa prices were leading companies such as Hershey, Nestlé, and Mars to reduce chocolate content, rely more on fillers, and send smaller quantities of chocolate per bag while keeping sticker prices similar—classic shrinkflation visible to shoppers.(fooddive.com)

While some analysts emphasized that the steepest retail price impact would arrive in late 2024 and into 2025, these reports show that within months of March 2024 consumers were indeed encountering smaller chocolate products and higher effective prices from major brands, driven by the cocoa price spike. That matches the substance of Friedberg’s prediction.

As of late 2025, percentage-of-sale-price commissions remain the dominant way U.S. residential real estate agents are paid. Post–NAR settlement coverage shows that average total commissions are still around 5–6% of the sale price, with only modest declines in some data (e.g., from about 5.64% to 4.96% in one survey), and in some cases average percentage commissions have even ticked up slightly rather than down. (kiplinger.com) Buyer–agent agreements introduced after the 2024 settlement typically still specify compensation as a percentage of the purchase price rather than as flat fees. (washingtonpost.com)

Flat-fee and fixed-fee brokerages do appear to be growing and are frequently discussed as alternatives that can save consumers money, but coverage consistently presents them as niche or emerging options rather than the prevailing model. Articles from Bankrate, NPR, and others describe flat-fee brokers as new opportunities created by rule changes, contrasted against the longstanding norm of 5–6% percentage-based commissions; data and reporting do not show flat-fee models as a majority of transactions or the primary compensation structure. (bankrate.com)

Because Friedberg’s prediction was explicitly framed as a gradual structural shift occurring “over time” without a clear deadline, and such an industry-wide change would reasonably be expected (if it happens at all) over many years rather than within roughly 18–20 months, it is too early to say definitively that the prediction is right or wrong. What we can say is that the predicted shift has not yet occurred by 2025 (percentage-based commissions are still dominant), but the open-ended time horizon means the ultimate outcome remains unresolved. Therefore, the appropriate status is inconclusive (too early) rather than definitively right or wrong.

Available data show only a modest decline in National Association of Realtors (NAR) membership so far, far short of the 500,000–1,000,000 exits Friedberg predicted, but his timeframe (“over time,” “over the coming years”) is too vague and clearly multi‑year, so it’s premature to call the prediction right or wrong.

What’s happened so far

• NAR ended 2023 with about 1,554,604 members, down roughly 26,000 (‑1.7%) from a year earlier. (wrenews.com)
• As of Oct. 31, 2024, NAR reported 1,526,631 members, still “near its all‑time high,” only ~2% below 2023 year‑end. (nar.realtor)
• A January 2025 tally based on state counts put membership at about 1.498 million, the lowest since March 2024 but still close to 1.5M. (realestatenews.com)
• At NAR’s June 5, 2025 board meeting, the treasurer reported 1,453,690 members as of May 31, 2025—a drop of roughly 100k–120k from late 2023, and still above NAR’s own 1.4M forecast for 2025. (nar.realtor)

So, roughly 100k–150k net members have left since the pre‑settlement peak—well below the 500k–1,000k contraction Friedberg anticipated.

Why it’s still too early to judge

• Friedberg talks about changes playing out “over time” and “over the coming years,” not by 2025 specifically. The core structural shifts from the commission settlement are only partly in effect: some practice changes started in 2024, and a key rule change—NAR no longer recommending that MLSs require NAR membership—doesn’t take effect until the 2026 MLS handbook, which could accelerate exits later in the decade. (barrons.com)
• Given that the horizon is implicitly multi‑year and the main post‑settlement membership shocks (like optional MLS‑NAR linkage) are still ahead, current data neither confirm nor definitively refute a long‑run 500k–1M decline.

Because the observed drop so far is much smaller than predicted but the stated timeframe has clearly not run its course, the fairest status as of November 30, 2025 is “inconclusive (too early).”

Parts of the prediction materialized, but the core claim about meaningful near‑term adoption that reduces or replaces traditional agents has not.

1. Startups did appear quickly and fit the described model.
After the March 2024 NAR commission settlement and related rule changes (implemented August 17, 2024), multiple startups launched or repositioned specifically around AI‑enabled, direct‑to‑consumer home‑buying tools:

• reAlpha launched an AI-powered, commission‑free “Super App” with Claire, a generative‑AI buyer’s agent, explicitly timed to the NAR rule changes and pitched as an end‑to‑end alternative to traditional buyer’s agents, including AI document review, negotiation guidance, and human backup when needed. (stocktitan.net)
• Modern Realty (YC S24) offers an AI real estate agent that guides buyers through the process, including disclosure summaries, offer generation, and scheduling showings; it frames itself as helping people buy homes “without relying on a traditional realtor,” though humans still review key steps. (ycombinator.com)
• Homa targets unrepresented buyers, using AI to generate CMAs, fill out state‑specific forms, and coach negotiations; it’s explicitly designed for buyers who want to navigate without an agent. (housingwire.com)
• Joy AI, RealStar AI, reAlpha, Ridley, and others market AI “real estate agents” or AI‑heavy, low‑ or no‑commission models that handle search, disclosure/document review, and negotiation support, often citing the NAR/DOJ changes as part of their rationale. (reddit.com)
• A 2025 Business Insider roundup highlights eight post‑settlement startups (including Homa and Ridley) using AI or other tech to cut commissions and reduce reliance on traditional agents. (businessinsider.com)

So the “a lot of startups will show up fast with AI, a‑la‑carte, DTC services” component is directionally correct.

2. But traditional agents remain overwhelmingly dominant, with little evidence of “meaningful adoption” of AI DTC alternatives.

• NAR’s 2025 Profile of Home Buyers and Sellers (covering July 2024–June 2025, i.e., post‑settlement) reports that 88% of buyers purchased through an agent or broker and 91% of sellers used an agent—matching or exceeding prior highs. FSBO sales are at or near record lows (~5–6%). (nar.realtor)
  This indicates that, at the market level, buyers and sellers have not shifted away from traditional agents in any large way.
• Legal/industry analyses note that commissions have barely moved and that agents often arrange commission splits informally, blunting the rule changes. A RISMedia/ABA discussion finds early “workaround” startups that advertised commissions outside the MLS “did not catch on.” (rismedia.com)
• NAR‑cited data for 2023–2025 shows buyers’ reliance on agents actually increasing slightly vs. pre‑settlement years (roughly mid‑80s% up to high‑80s%), which is the opposite of a rapid shift toward AI‑only services. (houstonagentmagazine.com)

3. Current AI real‑estate platforms are niche in scale and usually augment agents rather than replace them.

• The HousingWire feature profiling reAlpha, Modern Realty, and Homa explicitly frames these as early, limited‑geography platforms (e.g., reAlpha only in Florida, Modern Realty only in California; Homa just launched and is bootstrapped), with humans still involved for offers and negotiations. (housingwire.com)
• Linkhome/HomeGPT, an AI‑driven brokerage active since before the settlement, processed about $48.6M in 2024 transaction volume vs. roughly $1.5T in total U.S. home sales—on the order of 0.003% of market volume, underscoring that even one of the “fastest‑growing AI real estate platforms” is still tiny at national scale. (streetinsider.com)
• Market research on “AI in real estate” finds growth primarily in tools used by agents (lead gen, pricing, marketing), with the real estate agent segment holding the largest share of AI‑related usage—again indicating augmentation of agents rather than mass consumer migration to agent‑free AI platforms. (market.us)

4. Timing: the “near term” window has passed without clear displacement of agents.
The prediction was made in March 2024 about effects “coming out of this ruling” and “show[ing] up real fast.” By late 2025:

• NAR data shows no material drop in buyers’ use of agents; if anything, usage ticks slightly up, and FSBO is at record lows. (nar.realtor)
• Multiple consumer‑facing articles on the settlement emphasize that average commission rates and agent usage have barely changed, citing inertia and limited consumer awareness. (kiplinger.com)
• AI‑driven platforms are still described as emerging, “looking to find their stride in 2025,” and mostly confined to a handful of states. (loyaltyhomegroup.com)

Conclusion
The narrow part of the forecast—a wave of AI‑enabled, DTC, a‑la‑carte real estate startups appearing quickly after the NAR settlement—is supported by the evidence. But the crucial part—that these offerings would be compelling enough, in the near term, to meaningfully reduce or replace the need for traditional residential agents—has not happened. The overwhelming majority of buyers and sellers still use agents at historically high rates, commissions have barely shifted, and AI platforms remain niche and often agent‑augmented.

On balance, the prediction as stated is wrong.

Since the March 22, 2024 podcast, there has been clear early progress toward using genetically modified pig organs in humans, but these procedures remain highly experimental and far from routine clinical practice.

Key developments:

• Massachusetts General Hospital performed the first transplant of a genetically edited pig kidney into a living human in March 2024 and a second such transplant in January 2025; by early 2025, this was only the fourth pig-kidney recipient worldwide, underscoring how rare the procedure still is. (hms.harvard.edu)
• In February 2025, the U.S. FDA approved the first formal clinical trials of gene‑edited pig kidneys for patients with end‑stage kidney disease, with plans for up to ~50 participants across studies by United Therapeutics and eGenesis. (scientificamerican.com)
• By November 2025, NYU Langone had only just begun enrolling and operating on patients in one of these kidney xenotransplant trials. (news-medical.net)
• Additional experimental uses—such as FDA‑approved trials of gene‑edited pig livers used externally as a dialysis‑like bridge for liver failure, and pig lung transplants into brain‑dead donors for research—also remain investigational rather than standard care. (apnews.com)

These steps support the possibility that genetically modified animal organs could become a mainstream solution in the longer term, but they do not yet demonstrate that such organs are on a clearly established path to routine clinical use. Given that (1) only a handful of patients worldwide have received such organs in living clinical settings, (2) trials are just beginning, and (3) large‑scale safety, efficacy, regulatory approval, reimbursement, and ethical questions remain unresolved, it is too early in November 2025 to say whether the prediction that this will "definitely" become a widely adopted, mainstream medical solution is ultimately right or wrong.

Therefore, the appropriate judgment at this time is inconclusive (too early to tell).

As of November 30, 2025, there is strong early evidence in the direction of Friedberg’s thesis, but the claim is explicitly about where future AI-driven productivity and value will ultimately accrue, and the time horizon (likely late‑2020s/2030s) has not yet arrived.

1. Are traditional industrial sectors seeing meaningful AI-driven gains yet?
Yes, especially in manufacturing and pharma, but these gains are still in the early scaling phase rather than a settled end state:

• Manufacturing: Recent industry surveys report very high AI adoption and large reported impacts—e.g., up to ~63% productivity boosts and ~20% cost reductions in plants that have implemented AI at scale, with widespread use in predictive maintenance, visual inspection, and smart supply chains. (zipdo.co) Other data show AI integrated into many factories’ robotics and control systems, with projections of ~20% productivity growth in manufacturing by 2035 and large cost savings from predictive maintenance. (gitnux.org) Most manufacturers say AI is already improving quality and decision‑making and expect it to materially boost revenue over the next few years. (manufacturingtomorrow.com)
• Medicine/pharma: Think‑tank and consulting reports find AI can plausibly cut drug‑development timelines roughly in half and add 2.6–4.5% of annual revenues in pharma and medical products, with pharma companies currently capturing about half of the generative‑AI‑in‑drug‑discovery market’s revenue. (itif.org) Large incumbents like Eli Lilly are now building AI supercomputing partnerships specifically to accelerate discovery and manufacturing, reflecting a serious bet that AI will materially improve productivity and time‑to‑market. (reuters.com)
  Overall, this supports the direction of his view that traditional industrial sectors can get large productivity gains from AI, software, robotics, and automation, but much of the quantified impact is still framed as 2030+ potential rather than already realized.

2. Is “much of AI’s economic value” clearly accruing to these sectors instead of pure‑tech companies yet?
Here the picture is mixed and clearly not settled:

• On the one hand, broad surveys show that only a small minority of firms globally (about 5%) are getting measurable value from AI so far, and the industries with the highest AI maturity and realized value today include software, telecom, and fintech, while many traditional sectors (e.g., chemicals, construction, real estate) lag. (businessinsider.com) This suggests that, at least in the 2023–2025 window, realized AI value is still concentrated in tech and a handful of more advanced adopters rather than being clearly dominated by traditional industrial incumbents.
• Equity‑market evidence shows that the largest and most obvious financial beneficiaries of the AI boom to date are mega‑cap technology and semiconductor companies like Nvidia, Microsoft, Alphabet, and others, whose market caps have surged to multi‑trillion‑dollar levels largely on AI infrastructure and platform economics. (euronews.com) At the same time, many AI‑native startups are struggling to convert hype into revenue, so value within tech is skewed to large incumbents rather than new entrants, but it is still overwhelmingly tech‑sector value, not clearly dominated by food, manufacturing, or pharma stocks. (investor.wedbush.com) Because Friedberg’s prediction concerns where the bulk of AI’s economic value will ultimately accrue—a structural, long‑run outcome—not just who is winning in the first 1–2 years of the boom, current data on market caps and early adoption do not yet suffice to judge him right or wrong.

3. Why this is “inconclusive (too early)” rather than right/wrong/ambiguous

• Too early relative to the horizon: The core of the claim is explicitly forward‑looking (“will derive substantial future gains” and “that’s where I’d concentrate capital”). Most rigorous analyses of AI in manufacturing and medicine frame the main productivity and economic impact in the late‑2020s and 2030s, not by 2025. (gitnux.org)
• Evidence is directionally supportive but not decisive: We already see meaningful pilot‑ and early‑scale productivity gains in industrial sectors, consistent with his thesis, but they do not yet amount to clear evidence that these sectors are capturing most of AI’s economic value compared with tech platforms.
• Outcome metrics are inherently long‑run: Determining whether "much" of AI’s eventual economic value accrues to traditional industrial incumbents versus pure‑tech firms will require observing several more years of diffusion, commoditization of core models, and integration into physical production and healthcare systems.

Because both the productivity gains and the cross‑sector value distribution he describes are still unfolding on a longer time frame than the ~20 months since the podcast, the prediction cannot yet be fairly scored as right or wrong; it is best classified as inconclusive (too early).

There is not yet enough elapsed time or clear market evidence to judge this long‑term consolidation prediction.

The claim is that “over the coming years of the AI era” each major professional‑services vertical (law, accounting, creative, etc.) will see market share concentrate into one leading vertical AI provider that becomes the default service (a de‑facto “one lawyer service,” “one accounting service,” etc.). That is a multi‑year, possibly decade‑scale forecast.

As of late 2025, the evidence we do have shows:

• Legal AI is still highly fragmented. Multiple heavily funded startups and incumbents compete—e.g., Harvey (used by many top US law firms), Luminance, Legora, Eudia, Spellbook, Paxton, Theo, Marveri, and others—without a single provider clearly dominating all legal work or behaving as a universal “one lawyer service.” (businessinsider.com)
• Accounting AI likewise has several major players. Intuit’s QuickBooks/Intuit Assist is prominent and has been recognized as a leading AI accounting assistant, but it competes with Xero, Sage (Copilot), SAP (Joule), and others adding comparable AI functionality, again with no single universal monopoly‑like provider. (prnewswire.com)
• Major vendors are only now rolling out broader AI‑native accounting platforms (e.g., Intuit Accountant Suite launched in October 2025), which suggests we are still in an early competitive build‑out stage rather than a post‑consolidation equilibrium. (quickbooks.intuit.com)

However, the prediction is explicitly about what will happen "over the coming years" of the AI era, not by 2025. Current fragmentation neither proves nor disproves the future convergence Friedberg described; it only shows that consolidation has not yet occurred.

Because:

1. The time horizon is vague but clearly longer than ~1.5 years from March 2024 to November 2025; and
2. Present market structure (many competing vertical AI offerings, no clear single‑provider dominance per vertical) does not rule out future consolidation,

the correct assessment today is “inconclusive (too early)”, not right or wrong.

As of November 30, 2025, the U.S. Congress has not passed any federal law that (a) explicitly legalizes and broadly permits cultivated (lab-grown) meat nationwide and (b) expressly preempts state bans like Florida’s.

Instead, the federal landscape looks mostly unfavorable to cultivated meat:

• At the federal level, multiple bills have been introduced to restrict support for cell‑cultured meat. The REAL Meat Act of 2024 (H.R. 8757) and its successor, the REAL Meat Act of 2025 (H.R. 1116), seek to prohibit the use of federal funds for research, promotion, production, or federal program purchases of cell‑cultured meat, not to legalize or protect it from state bans. (davidson.house.gov) Another bill, the School Lunch Integrity Act, aims to keep cell‑cultivated meat out of federal school meal programs. (rounds.senate.gov) None of these are broad legalizations or preemption statutes, and none have become law in a way that overturns state bans.
• Meanwhile, more states have enacted outright bans on the manufacture or sale of cultivated meat, moving in the opposite direction of the prediction. Florida (SB 1084), Alabama (SB 23), and Mississippi have all passed bans, and by mid‑2025 Texas joined them with SB 261; analyses list additional states such as Indiana, Montana, Missouri, and Nebraska among those with bans or strong restrictions. (aglaw.psu.edu)
• The legal challenges to Florida’s ban are proceeding under existing federal meat laws (e.g., Federal Meat Inspection Act, Poultry Products Inspection Act) and constitutional doctrines like the Dormant Commerce Clause and Supremacy Clause, but no new federal statute has been enacted that clearly and expressly preempts state cultivated‑meat bans. Judges in the Florida litigation have explicitly noted that current federal approvals for cultivated chicken do not themselves stop states from banning these products. (health.wusf.usf.edu)

However, Friedberg’s time window was “over the next couple of years” from a March 15, 2024 release date, which reasonably extends to around March 2026. Because that horizon has not yet elapsed, it is too early to say with certainty that Congress will not enact some preemptive, pro‑cultivated‑meat statute before then, even though current legislative trends run against his scenario.

Given that the deadline he implied is still in the future and no definitive contrary event (such as a constitutional ruling locking in state authority to ban cultivated meat regardless of future federal law) has occurred, the prediction cannot yet be judged definitively right or wrong. It is therefore best scored as inconclusive (too early to tell).

Texas and several other states have, in fact, enacted bans similar to Florida’s on cultivated (lab-grown) meat.

• Florida (trigger state in the quote): SB 1084 created Fla. Stat. §500.452, making it unlawful to “manufacture for sale, sell, hold or offer for sale, or distribute cultivated meat” in the state, with criminal penalties. (leg.state.fl.us)
• Alabama: Act 2024‑252 makes it illegal in Alabama to manufacture, sell, or distribute food products produced from cultured animal cells, effective October 1, 2024. (alabamaretail.org)
• Mississippi: HB 1006, passed in March 2025, prohibits anyone from manufacturing, selling, holding for sale, offering for sale, or distributing any cultivated food product in the state, effective July 1, 2025. (food-safety.com)
• Indiana: HB 1425 (signed May 6, 2025) imposes a two‑year moratorium, from July 1, 2025 through June 30, 2027, on the sale, offer for sale, or manufacture of cultivated meat products in Indiana. (aglaw.psu.edu)
• Montana: HB 401, signed May 1, 2025, bans the sale, distribution, and manufacture of “cell‑cultured edible products,” with criminal penalties and license consequences starting October 1, 2025. (dailyintakeblog.com)
• Nebraska: LB 246 (approved May 20, 2025) amends the Nebraska Pure Food Act so that cultivated‑protein food products are treated as adulterated food; it makes their manufacture, production, import, distribution, promotion, display, offer for sale, attempt to sell, or sale unlawful statewide. (nationalaglawcenter.org)
• Texas (explicitly named in the prediction): SB 261, signed by Governor Greg Abbott on June 20, 2025, prohibits the manufacture, processing, possession, distribution, offer for sale, and sale of “cell‑cultured protein” in Texas, effective September 1, 2025. (capitol.texas.gov)

A legal and policy survey from agricultural law and food‑law analysts similarly lists Texas alongside Alabama, Florida, Indiana, Mississippi, Montana, and Nebraska as states that have enacted bans or moratoria on cultivated meat products. (aglaw.psu.edu)

Given that Texas implemented a ban and multiple other U.S. states subsequently enacted laws barring the manufacture and/or sale of cultivated meat, Friedberg’s prediction that Florida’s move would be followed by Texas and “many other states” adopting similar bans has come true.

Available usage data suggest that Apple’s global iOS share has been roughly flat or slightly down since the prediction, but the forecast explicitly referred to gains "over time" / "over the coming years," so it’s too early to judge it as clearly right or wrong.

What the data show so far

• Around the time of the episode (Q1 2024), StatCounter-based summaries put global mobile OS share at roughly 70–71% Android vs ~28% iOS (e.g., Q1 2024: Android 70.71%, iOS 28.54%). (soax.com)
• For 2024 as a whole, multiple StatCounter-derived sources show iOS in the 27–28% range (Android roughly 71%). (goodfirms.co)
• By 2025, StatCounter’s global mobile(-and-console) OS data and derivatives generally show Android ~72–73% and iOS ~27% (e.g., October 2025: Android 72.56%, iOS 26.99%; year-aggregated figures: Android ~72.5%, iOS ~27.0–27.5%). (gs.statcounter.com)

So, compared to the ~27–28% iOS share at the time of the prediction, today’s ~27% level does not yet reflect the clear, sustained "stealing" of share from Android that Friedberg anticipated.

Why this is still inconclusive

• The prediction was framed as a multi‑year structural story (rising GDP per capita in emerging markets + iOS remaining the superior experience → iOS gradually gaining share). That sort of shift would reasonably be evaluated over a horizon of several years, not ~1.5–2 years.
• Over just this short window, market share has been essentially flat with small statistical noise; that’s not strong evidence for or against a longer‑term directional call.

Given the vague time horizon ("over time," "over the coming years") and the modest elapsed time since March 2024, the fairest assessment is that the prediction’s accuracy cannot yet be definitively determined.

Within roughly a year of March 1, 2024, there was a clear shift from mostly unpriced, implicit scraping of web content toward a recognizable market for training data, with repeated, high-profile licensing deals that put concrete dollar values and contractual patterns around content used for large language models.

From spring 2024 onward, OpenAI and other AI firms signed a series of explicit, paid content-licensing agreements with major news organizations and platforms: Le Monde and Spain’s Prisa Media in March 2024; the Financial Times in April 2024; and a large, multi‑year News Corp deal in May 2024 that press reports said could exceed $250 million over five years, following earlier agreements with the Associated Press and Axel Springer. These contracts typically cover both training use and in-product display, and industry reporting indicates that OpenAI and peers now routinely pay anywhere from low single‑digit millions per year for smaller publishers up to nine‑figure packages for marquee groups, establishing de facto price bands for premium text data. (openai.com) In parallel, platform data deals such as Reddit’s reported $60 million‑per‑year agreement with Google for API access, plus subsequent AI partnerships, reinforced the notion that large conversational datasets have substantial, quantifiable market value. (time.com) Broader industry analysis by mid‑ to late‑2024 described a “data gold rush” in which big tech firms were systematically hedging legal and supply risk by buying or licensing training data (e.g., Shutterstock image/video/music libraries, Defined.ai brokered datasets) at increasingly standardised per‑asset or per‑word prices, further clarifying commercial norms for AI training data. (wifc.com)

At the same time, the legal framework around copyright and fair use for training remained unresolved, with major lawsuits such as The New York Times and other publishers vs. OpenAI and Microsoft allowed to proceed into discovery in 2024–2025 rather than being definitively settled, and new legislation like the U.S. Generative AI Copyright Disclosure Act only beginning to address transparency obligations. (apnews.com) That ongoing litigation shows the law is not fully settled, but it does not negate the prediction: by early 2025, there was a much clearer, widely reported pattern of how AI companies and rights‑holders were valuing and transacting over training data (license-or-litigate, with well‑understood deal structures and price ranges), compared with the far murkier situation in early 2023–early 2024. Given that the claim was that things would start to become clearer over the following year, not that every legal and economic question would be fully resolved, the prediction is best judged as having come true.

Major general‑purpose AI Q&A services have become commercially successful without offering the kind of explicit user‑level values and filtering personalization Friedberg described.

1. Strong commercial success of non‑personalized LLMs

   • OpenAI’s ChatGPT has on the order of 700–800 million weekly active users in 2025, processes billions of messages per day, and is generating revenue at a run rate in the tens of billions of dollars. (business-standard.com)
   • Google’s Gemini app has also grown rapidly, surpassing 450M+ monthly active users by mid‑2025 and then around 650M MAUs by Q3 2025. (timesofindia.indiatimes.com)
     These are clear cases of large‑scale commercial success.

2. What personalization actually exists

   • ChatGPT offers Custom Instructions and “Memory mode,” letting users set preferences for tone, level of detail, language, and some persistent context across chats—essentially style and task personalization. (tomsguide.com)
   • Google Gemini similarly added Personal Context / memory so it can remember user details and preferences to tailor answers, plus features like Temporary Chats. (theverge.com)
     These are convenience and style controls, not user‑selectable moral or political value systems or per‑user safety/"wokeness" levels.

3. Core values and filtering remain centrally set, not user‑personalized

   • Anthropic’s Claude is governed by a fixed “Constitutional AI” framework: a single set of principles (e.g., human‑rights‑inspired rules) that define model behavior, explicitly not per‑user moral personalization. (anthropic.com)
   • Mainstream U.S. chatbots—including ChatGPT, Gemini, Copilot, and Claude—still enforce unified safety and content policies around politics, hate speech, and harmful instructions. Regulatory and political pressure (e.g., U.S. state investigations into alleged political bias) has pushed toward tighter, centralized controls, not optional user‑controlled value dials. (theverge.com)
     In other words, the industry did not move to broad user‑level value/filter personalization as a prerequisite for success.

4. Users do encounter objectionable or misaligned outputs—yet overall usage grows

   • Google’s Gemini image‑generation controversy in February 2024 is a clear example where outputs (racially anachronistic historical images, odd moral answers) offended and angered users; Sundar Pichai publicly acknowledged Gemini’s responses had "offended our users" and showed bias, and Google temporarily paused people‑image generation. (wusf.org)
   • Anecdotally, some users say they “don’t use Gemini anymore” or complain that Gemini or Grok became too censored or too "woke" and claim to have dropped them. (reddit.com)
   • However, despite repeated controversies, both ChatGPT and Gemini show rapidly rising user numbers and revenues, not broad abandonment. (business-standard.com) Surveys of AI attitudes show growing concern and mistrust around AI, but mainly about jobs, misinformation, and general unease—not documented mass churn specifically because outputs clashed with individual values. (theverge.com)

5. Net assessment vs. the prediction

   • Friedberg’s strong claim was that the only way such Q&A services would be successful is by enabling user‑level value/filter personalization, and that if they didn’t, users encountering misaligned outputs would broadly say "I don’t want to use this product anymore."
   • In reality, by late 2025, the leading systems:
     • Have limited, mostly stylistic/context personalization, not user‑selectable moral or ideological filters.
     • Have faced real backlash over perceived bias and "wokeness," and some individual users have indeed quit.
     • Yet are commercially thriving at very large scale, with usage and revenue still growing.

Because general‑purpose AI Q&A products have become widely and commercially successful without the kind of user‑level value/filter personalization he described, and because there is no evidence that a "significant share" of users has abandoned them for that specific reason, the core conditional claim is contradicted by observed outcomes. Hence the prediction is best classified as wrong, even if it correctly anticipated that alignment controversies would matter for some users and spur demand for more personalization features.

The prediction had two parts: (1) many competing models, including open source, would proliferate, and (2) open‑source models would gain a significant competitive advantage and effectively ‘win’ over tightly controlled proprietary models. The first part is clearly happening: surveys and market analyses show that more than half of organizations now use open‑source AI tools or models alongside proprietary ones, and open‑source ecosystems like Llama, Qwen, Mistral, and DeepSeek are a major driver of innovation and cost savings.(mckinsey.com) However, the second, stronger claim that open source would win in the market has not materialized by late 2025. A recent study (Nagle & Yue) finds open models account for only about 20% of usage and roughly 4% of AI-market revenue, despite similar performance and much lower operating cost, indicating that revenue and overall usage are still dominated by closed models.(itpro.com) Enterprise data from Menlo Ventures and related analyses show proprietary providers like Anthropic, OpenAI, and Google now command the vast majority of enterprise LLM workloads, with Anthropic at ~32%, OpenAI ~25%, Google ~20%, while open‑source models (largely Llama and a few others) serve only about 13% of corporate workloads and that share has recently declined.(finance.yahoo.com) Commentaries on the 2025 enterprise landscape explicitly note that open‑source adoption has plateaued or fallen in production use, with organizations sticking to closed models due to performance gaps, integration complexity, and compliance concerns, even though they recognize open source’s cost advantages.(itpro.com) In short, while dissatisfaction with heavily filtered models did help spur a diverse, vibrant open‑source ecosystem, the market outcome so far is a hybrid world where closed models dominate spend and mission‑critical workloads; open source has grown and is important, but it has not secured the clear competitive market win the prediction envisioned.

Friedberg expected that within a few months of February 2024, LLM quality would advance so fast that early‑2024 models would soon look primitive and notably bad, like 1996 web search. In reality, the main releases in the next few months—Google’s Gemini 1.5 (announced Feb 15, 2024), Anthropic’s Claude 3 family on March 4, 2024, and OpenAI’s GPT‑4o on May 13, 2024—were meaningful but incremental improvements over GPT‑4, Claude 2.x, and Gemini 1.0, not an order‑of‑magnitude jump. Claude 3 and later 3.5 Sonnet slightly outperform GPT‑4 on many benchmarks, while GPT‑4o raises MMLU from about 86.5 to 88.7 and improves multimodal and voice capabilities, but these are refinements rather than a shift that makes earlier models look ‘bad’.(en.wikipedia.org) Even by late 2025, industry coverage characterizes ChatGPT’s evolution through GPT‑4o, GPT‑4.5, and GPT‑4.1 as steady, incremental progress with ongoing issues like hallucinations, rather than a single transformative break point.(techradar.com) Older models such as GPT‑4 and Claude 3 remain in active use and are still viewed as strong systems, not as obviously obsolete in the way mid‑1990s web search is compared with later generations.(help.openai.com) Thus, both the short timeline (‘a couple of months’) and the implied scale of quality change did not materialize as predicted.

Available evidence to late 2025 shows no meaningful user exodus from Google Search or its broader consumer ecosystem due to the Gemini/“woke AI” controversies, and Google has maintained – even strengthened – its dominance.

1. Gemini “woke” backlash happened, but Google course‑corrected rather than simply “continuing on the same path.” After the February 2024 uproar over racially inaccurate historical images and other politically skewed outputs, Google paused Gemini’s ability to generate images of people and publicly acknowledged it had “got it wrong,” promising and implementing fixes. Later controversies (e.g., on how Gemini discussed pedophilia) added criticism but were again met with adjustment and public contrition, not a hardline insistence on the original behavior. (forbes.com) This weakens the antecedent of Friedberg’s conditional claim (“if Google continues on this path”).

2. Search users have not shifted in large numbers to competitors; Google’s market share remains ~90%. As of May 2025, StatCounter‑summarized data shows Google at roughly 89–90% global search share, with Bing at ~4% and other search engines in the low single digits. This is the first time in over a decade Google even dipped slightly below 90%, but it remains overwhelmingly dominant in both global and U.S. search. (en.wikipedia.org) There is no sign of a mass user migration that would constitute a “meaningful” loss of users in the sense Friedberg implied (i.e., enough to seriously erode dominance).

3. Google’s search and ads businesses are growing, not shrinking, despite AI competition. Alphabet’s 2024–2025 earnings show double‑digit revenue growth and strong search/ads performance: Q4 2024 and multiple 2025 quarters report Search & Other ad revenue up roughly 10–15% year‑over‑year, with total advertising revenue also growing solidly. (cnbc.com) A Nasdaq analysis of Alphabet’s “$100 billion quarter” notes that Search still holds about 90% of global search traffic and that AI‑driven search features are supporting, not undermining, ad revenue. (nasdaq.com) This is inconsistent with a story where market forces are punishing Google by pulling users and ad dollars away.

4. Google’s AI features are seeing massive adoption rather than abandonment. AI Overviews in Search reached over 1.5 billion monthly users by Q1 2025, and Google reports hundreds of millions of active Gemini users, with rapid growth. (cnbc.com) That suggests users are engaging with Google’s AI‑infused search rather than fleeing it for rivals.

5. Rivals like Perplexity and Bing are growing, but at a scale that doesn’t yet threaten Google’s dominance. Perplexity processed about 780 million queries in May 2025 (~30 million per day) and has raised large sums at a high valuation, but that volume is tiny compared to Google’s billions of daily searches. (en.wikipedia.org) Bing’s global share remains only a few percent. (en.wikipedia.org) These gains are meaningful for startups and Microsoft but not yet “market correction” on the scale Friedberg described.

6. Financial markets view Alphabet as an AI and search winner, not a company being disciplined for an inferior product. Reporting from the FT, WSJ, Reuters, Barron’s, and others in late 2025 describes Alphabet adding over $1 trillion in market cap, nearing a $4 trillion valuation, and being treated as a leading AI beneficiary. Analysts emphasize its reinforced dominance in search and AI and its strong earnings and margins. (ft.com) That outcome is the opposite of what you’d expect if “the market [were] not going to fail us” and were punishing Google for keeping an inferior, ideologically skewed product.

Putting this together: while there was real user backlash and some incremental competitive pressure, there is no strong evidence that Google has lost a meaningful number of users/consumers in a way that seriously erodes its search dominance or that market forces have “corrected” Google’s position. Instead, Google remains the overwhelmingly dominant search provider with growing AI usage and strong financial performance. On the observable 2024–2025 data, Friedberg’s prediction has not materialized and is best scored as wrong rather than merely “too early” or ambiguous.

As of November 30, 2025, no single company has achieved anything close to a monopoly on AI training data or on informational control via paid content licensing.

On the data supply side, there is an active, multi‑firm race to license training data: OpenAI, Google, Meta, Microsoft, Apple, Amazon and others all buy datasets from publishers, stock-image libraries, and brokers, rather than one firm dominating the market. Reuters describes a "generative data gold rush" in which all of these companies are licensing archives and specialized content, not just a single player. (reuters.com) OpenAI has signed numerous media deals (Associated Press, Axel Springer, News Corp, Future, Vox, The Atlantic, Axios, etc.), with at least one major deal (News Corp) reported as exclusive for that publisher’s journalism, but most others are explicitly non‑exclusive and confined to particular archives. (apnews.com) This gives OpenAI advantages, but not a monopoly over all high‑value human text.

Crucially, the open internet and open corpora remain very large and widely used. Common Crawl alone provides hundreds of billions of web pages as a free, open corpus used in many LLMs. (commoncrawl.org) The Allen Institute’s Dolma dataset (≈3T tokens) is fully open and powers the OLMo family of models, with both data and code released. (allenai.org) Harvard’s public‑domain books corpus and open datasets like The Pile further expand freely available high‑quality text. (wired.com) Competitive open‑ or semi‑open models such as Meta’s Llama 3, DBRX, and others, trained heavily on public data, demonstrate that state‑of‑the‑art systems can still be built without relying on one firm’s licensed archives. (techcrunch.com) Regulators and researchers do worry that large firms could use data and partnership "moats" to entrench power, and that crawler blocking by many news outlets (especially high‑factual ones) may skew future training data. (time.com) But these are warnings about potential concentration, not findings that a single actor already controls information.

Because Friedberg’s claim is forward‑looking ("there isn’t going to be a monopoly" and we won’t be "monopolized away into some disinformation age"), it cannot be definitively verified or falsified after only ~21 months. What can be said is that, so far, the world looks more like his scenario than its opposite: multiple powerful firms, robust open data sources, and no single company with effective monopoly control over training data or information. However, future legal, economic, or regulatory shifts could still change this, so the prediction’s ultimate truth value remains inconclusive (too early) rather than clearly right or wrong.

Friedberg's claim was that the next major iPhone and flagship-Android cycle after February 2024 would ship with chips explicitly designed to run AI models locally and enable substantial on‑device AI features. That is what happened.

Apple's iPhone 16 lineup with A18 and A18 Pro chips (launched September 2024) is marketed as being built for Apple Intelligence; Apple states that many of the generative models behind Apple Intelligence run entirely on‑device, and that third‑party apps can call those on‑device models via the Foundation Models framework, with features that can work offline. (apple.com) Apple and press materials also describe the A18 Neural Engine as optimized for large generative models and significantly faster at running ML workloads, specifically to accelerate Apple Intelligence, and note that Apple Intelligence requires newer A‑series chips with sufficient GPU/NPU capacity for on‑device processing. (businesswire.com)

On Android, Samsung's Galaxy S25 series uses Qualcomm's Snapdragon 8 Elite for Galaxy, whose substantially faster NPU is highlighted as enabling many more Galaxy AI experiences to run on‑device, including features like Generative Edit that previously relied on the cloud; Samsung also exposes a setting to disable cloud AI and rely only on local processing. (techpowerup.com) Google's Pixel 9 family, powered by the Tensor G4 chip, runs Google's Gemini Nano large language model locally; this on‑device LLM powers features such as Recorder summaries, Call Notes, Pixel Screenshots, and other real‑time summarization, translation, and captioning capabilities that operate without sending data to the cloud. (androidauthority.com) Other flagship SoCs like MediaTek's Dimensity 9400 and Samsung's Exynos platforms are likewise advertised as supporting on‑device generative AI, including multimodal Gemini Nano, underscoring that high‑end phones are now expected to run such models locally. (semiconductor.samsung.com)

Given that the post‑February‑2024 flagship cycles from Apple, Samsung, Google, and other major vendors did in fact ship with dedicated NPUs/AI engines specifically marketed for running sizable generative models on the device and enabling prominent on‑device AI features, Friedberg's prediction that the 'next set of mobile devices' would have chips that run models locally has been borne out.

Nvidia’s valuation has grown dramatically since the prediction, but it has not reached $10T, nor has anything happened that clearly makes a $10T outcome impossible.

• At the time of the podcast (February 2024), Nvidia was crossing the ~$2T mark; news coverage notes it hit $2T in February 2024 and then rose to about $2.25T by March 2024, up from roughly $1.2T at the end of 2023. (ndtv.com)
• As of late November 2025, Nvidia’s market cap is around $4.4T, making it the world’s most valuable company, and it has traded as high as roughly $4.5T in October 2025. (companiesmarketcap.com)
• The condition attached to the prediction—continued strength in AI hardware/efficiency trends—largely holds so far: Nvidia remains the dominant AI chip supplier, Blackwell-based GPUs are ramping strongly, and the roadmap to Rubin and beyond underpins very high expected AI demand and revenue (management and analysts discussing targets of hundreds of billions in AI-chip revenue through 2026). (cnbc.com)

However, the prediction was long term and framed as potential (“could be worth $10 trillion, potentially”), with no specific deadline. Nvidia has not yet reached $10T, but given that it is already in the mid‑trillions and AI hardware demand remains elevated, it is not possible to say that a $10T valuation has been disproven. It remains an open, speculative outcome rather than a resolved one.

Because the core claim is about a possible long‑term ceiling and no explicit timeframe has elapsed in which that ceiling would need to be hit or clearly ruled out, the status of “Nvidia could be worth $10T in the long term” is too early to judge.

Available evidence by November 30, 2025 does not clearly confirm or falsify Friedberg’s prediction.

• The precise scenario hasn’t really happened yet. Apple Vision Pro only launched in 2024 and, while Apple and partners highlight enterprise use cases (training, remote fieldwork, digital twins, industrial visualization), there is no public documentation of it being deployed in the specific greenhouse or lab‑technician workflows Friedberg described (QR codes on plants, task lists, in‑situ data capture) at scale, let alone with measured productivity multipliers. (apple.com)
• Closest real-world analogs (non‑Apple smart glasses in agriculture/livestock) are still pilot‑level and mostly qualitative. Multiple studies and trials show that AR smart glasses can support QR‑code based data access, remote assistance, and hands‑free operation in farms and livestock settings, including detailed measurements of QR scanning performance. These papers and evaluations conclude that smart glasses are promising for improving workflow efficiency, but they do not report anything like a 10× increase in overall worker productivity versus phone‑ or paper‑based systems. (mdpi.com)
• Greenhouse-focused pilots exist but lack quantified 10× gains. Smart-glasses projects in greenhouses (e.g., Tokyo tomato greenhouses using 5G + smart eyewear, and smart glasses assisting cluster‑tomato harvesting by assessing ripeness) report that the devices make tasks easier or support operation with fewer or less‑experienced workers, but public write‑ups do not quantify a roughly 10× per‑worker productivity improvement versus previous phone/scanner/printer workflows. (natureworldnews.com)
• General XR/AVP productivity research is still early and mixed. Broader XR and Apple Vision Pro studies in professional contexts suggest some productivity and usability benefits, but also emphasize limited app ecosystems, integration friction with existing workflows, and user comfort constraints. These findings indicate that the long‑term productivity potential is not yet fully realized or measured in real industrial or agricultural deployments. (arxiv.org)
• Friedberg’s own statement is explicitly long‑term and conditional. He frames the 10× improvement as what will happen if AR goggles like Vision Pro are properly deployed with the necessary software and adoption in those workflows, not as something expected within months. Secondary coverage repeats his view as a forward‑looking thesis about spatial computing in agriculture, not as a claim that has already been validated. (agtecher.com)

Because (1) the specific Vision Pro/AR‑goggles greenhouse workflow Friedberg described has not yet been widely implemented with mature software, and (2) existing field and academic studies in adjacent agricultural/livestock scenarios do not provide clear, quantitative evidence of ~10× productivity gains, the prediction cannot currently be tested. It remains a plausible but unproven long‑term claim, so the outcome is inconclusive rather than clearly right or wrong.

The prediction is that Apple will generate at least $100 billion in cumulative revenue from Apple Vision Pro hardware (current and next‑generation models) within less than five years of launch, i.e., before around February 2029.

As of the current date (November 30, 2025):

• Fewer than two years have passed since the initial Vision Pro launch in early 2024.
• Public reporting and analyst estimates indicate Vision Pro sales are in the hundreds of thousands to low single‑digit millions of units, not tens of millions, and nowhere near $100B in revenue yet. At roughly $3,500 per unit, Apple would need on the order of ~28–29 million units sold to reach $100B ([100,000,000,000 ÷ 3,500 ≈ 28.6M], see calculation).
• There is no evidence that Apple has approached $100B in cumulative Vision Pro revenue so far; that would be an extraordinary, widely reported milestone if it had occurred.

However, the deadline for the prediction is February 2029, and that date has not yet arrived. The prediction could still come true if future Vision Pro generations sell much more strongly.

Therefore, it is too early to determine whether the prediction is ultimately correct or incorrect.

As of November 30, 2025, both the trigger and the time window in Friedberg’s conditional prediction have not clearly been met.

1. Office values have fallen, but not by “roughly two‑thirds.” Nationally, average U.S. office sale prices are down about 37% from 2019 levels as of the end of 2024, with forecasts of roughly another 10% decline over the rest of the decade, not a 60–70% collapse in aggregate values. (businessinsider.com) Academic work on the “office real estate apocalypse” estimates nationwide office value destruction of roughly $500–550B from 2019 levels—substantial, but far short of wiping out two‑thirds of the asset class if the total stock is around the low–single‑digit trillions. (business.columbia.edu) A Washington Post summary of the same research notes about $557B in office value lost from 2019–2023, again implying a drop on the order of tens of percent, not two‑thirds. (washingtonpost.com)

2. No targeted federal bailout for pensioners’ office losses has been enacted. Congress has held hearings on commercial real estate stress, where industry representatives explicitly said they were not seeking a bailout and instead asked for regulatory and accounting flexibilities. (marketwatch.com) There has been no new TARP‑style or Butch Lewis‑style program specifically designed to compensate retirees or pension plans for office‑sector write‑downs; existing pension assistance laws (e.g., the Butch Lewis Act’s multiemployer relief) predate this podcast and are not office‑specific. (en.wikipedia.org)

3. The time horizon (“within the following few years”) is not over. The episode aired in February 2024, so fewer than two years have passed—well short of a typical “few years” reading—and the extreme valuation‑collapse scenario that was supposed to trigger a bailout has not occurred.

Because the prerequisite scenario (a ~two‑thirds collapse in U.S. office values severely hitting pensions) has not fully materialized, and the envisaged multi‑year window for a policy response is still open, Friedberg’s conditional prediction cannot yet be judged right or wrong.

It is too early to determine whether this prediction about future multi‑decadal frequency of California‑scale ARkStorm‑type mega‑floods has come true.

Friedberg’s claim is about changing recurrence intervals: events that historically occurred roughly every 150–200 years purportedly shifting to every ~25–50 years or sooner, driven by warmer oceans and increased atmospheric river activity. That is, it is a statistical, long‑horizon prediction about the frequency and severity of extreme events over many decades.

As of the current date (November 30, 2025), only about one year and ten months have passed since the prediction (January 26, 2024). That time span is far too short to empirically confirm or refute a change in the recurrence rate of phenomena that are, by definition, multi‑decadal to centennial in their historical frequency. Even if one or more extreme storms occur in this short period, that would not be statistically sufficient to validate a claimed 25–50‑year recurrence pattern, nor would the absence of such events in <2 years disprove it.

Current climate research does support that:

• Atmospheric rivers impacting the U.S. West Coast are expected to increase in intensity and may increase in frequency as the climate warms, due in part to higher sea‑surface temperatures and increased atmospheric moisture.
• Studies and USGS/NOAA summaries of ARkStorm‑type scenarios emphasize that warming is likely to raise the risk of extreme precipitation and flooding events in California over the 21st century.

However, while this research is qualitatively consistent with the direction of Friedberg’s statement (more frequent/severe extreme precipitation events in a warmer world), the specific quantitative claim—that ARkStorm‑scale floods will reliably shift from ~150–200‑year events to ~25–50‑year events or less—cannot be validated or falsified based on observations in the very short period since the prediction was made. That would require multiple decades of data and/or a clear consensus in the literature on that exact new recurrence interval, which we do not yet have.

Because the prediction concerns a long‑term statistical frequency and only a short amount of time has elapsed, the correct classification is: it’s too early to tell whether it has come true.

Available evidence shows that in the ~two weeks after the January 26, 2024 recording (through about February 9, 2024), California did have very wet conditions, including major storms, but did not experience an ARkStorm‑ or ARkStorm 2.0‑level megaflood.

1. What an ARkStorm/ARkStorm 2.0 event would look like

   • ARkStorm is a hypothetical California “megastorm” scenario involving weeks of near‑continuous intense atmospheric river storms, up to ~10 feet of water in some areas, and estimated damages on the order of $725 billion to $1 trillion, far beyond typical severe storms. (nationalgeographic.com)
   • ARkStorm 2.0 is an updated modeling study showing climate change roughly doubles the odds of such a catastrophic megaflood, but it remains a scenario that has not yet occurred. (gmd.copernicus.org)

2. What actually happened in the couple of weeks after Jan 26, 2024

   • In early February 2024 (Feb 4–7), California was hit by two strong atmospheric river events, bringing record or near‑record rainfall, hurricane‑force wind gusts in some locations, widespread flooding, mudslides, power outages, and at least 9 deaths. (en.wikipedia.org)
   • Damage estimates for this storm sequence are on the order of billions of dollars (roughly $3 billion in direct damages and $9–11 billion in total economic loss), which is severe but still orders of magnitude below the ARkStorm megaflood scenario. (en.wikipedia.org)
   • National Geographic, writing on February 8, 2024, explicitly notes that while the first week of February saw “historic storms and record rainfall” that killed nine people and damaged numerous homes, this “pales in comparison” to an ARkStorm‑level event and that ARkStorm remains a hypothetical disaster being discussed in the context of social‑media rumors. (nationalgeographic.com)

3. Comparison to the prediction

   • Predicted: The next couple of weeks would be wet/very wet, but not an imminent ARkStorm 2.0‑style catastrophic megaflood.
   • Observed: Early‑February atmospheric rivers did indeed produce very wet conditions, historic rain totals, and serious but regional flooding and damage. However, by all available accounts they fell far short of the statewide, weeks‑long, trillion‑dollar, ARkStorm‑scale megaflood described in the scientific and emergency‑planning literature. (en.wikipedia.org)

Because California experienced intense storms and flooding but no ARkStorm‑level, ARkStorm 2.0‑style megaflood in the stated time window, Friedberg’s prediction is best judged as right.

Evidence from both futures markets in January 2024 and the subsequent Fed policy path shows Friedberg’s "higher for longer" call was directionally correct.

• What markets were pricing in January 2024:
– At the very end of 2023 and into early January 2024, futures and swaps markets were pricing six quarter‑point cuts in 2024, with the first cut expected in March, and at latest May.(business-standard.com)
– After the January 31, 2024 FOMC meeting, CME FedWatch–based analysis still implied the federal funds rate would fall to 3.75–4.0% by end‑2024, meaning roughly 150–175 bps of easing from the 5.25–5.5% range.(econbrowser.com)
– Commentaries at that time noted March cut odds had been near 90% before Powell pushed back, but markets still assumed a relatively early and aggressive cutting cycle.(rbcwealthmanagement.com)

• What actually happened:
– The Fed’s target range was raised to 5.25–5.5% in July 2023 and then held at that peak for more than a year, through mid‑September 2024.(axios.com)
– The first rate cut did not occur until September 18, 2024, when the FOMC reduced the range by 50 bps to 4.75–5.0%, explicitly described as the first cut in over four years.(axios.com)
– Subsequent 25 bp cuts in November and December 2024 brought the target range to 4.25–4.5% by year‑end, as confirmed by FOMC minutes and post‑meeting analyses.(federalreserve.gov)
– A March 2025 Federal Reserve Bank of San Francisco overview explicitly notes that the funds rate was cut by a cumulative 100 bps from its peak of 5.25–5.5%, with the target range at 4.25–4.5% by the January 2025 meeting.(frbsf.org)

• Comparison to the January 2024 futures path:
– Markets in January 2024 were effectively pricing:
1. First cut by March–May 2024; and
2. End‑2024 rate around 3.75–4.0%.(business-standard.com)
– In reality, the first cut slipped to September 2024, well after both March and mid‑2024, and the end‑2024 rate (4.25–4.5%) was materially higher than the 3.75–4.0% level implied by January futures.(axios.com)

• On his mechanism (lagging components): Later Fed commentary and research in early 2025 note that inflation in services such as shelter, health care, and insurance remained elevated and was a key reason policy stayed restrictive, which is broadly consistent with Friedberg’s argument that lagging components like insurance would keep inflation sticky.(frbsf.org)

Because (1) cuts were indeed delayed well past the March 2024 start and beyond mid‑2024, and (2) policy rates ended 2024 higher than what January 2024 futures were discounting, Friedberg’s prediction that U.S. rates would stay higher for longer than markets expected at the time is best judged as right.

Evidence from the 2024 election shows that no independent or third‑party effort became a “big winner” or materially challenged Democratic–Republican dominance.

• Final vote shares: In the November 5, 2024 presidential election, Donald Trump (R) won 49.8% of the popular vote and Kamala Harris (D) won 48.3%, with all 538 electoral votes going to the two major‑party tickets (312 for Trump, 226 for Harris).(en.wikipedia.org) Third‑party and independent candidates collectively received around 2% of the national vote, a slight uptick from 2020 but still marginal.
• Performance of leading third‑party/independent candidates: The strongest non‑major‑party campaigns each finished under 1% nationally: Jill Stein (Green) about 0.55%, Robert F. Kennedy Jr. (independent/We the People) about 0.48–0.49%, and Libertarian Chase Oliver about 0.42%.(en.wikipedia.org) None won any electoral votes or a substantial state‑level share comparable to historic disruptors like Ross Perot or George Wallace.
• RFK Jr.’s trajectory: Kennedy did initially poll unusually well for a third‑party hopeful, with analyses noting that his early polling was the strongest for such a candidate since Perot. But by August 23, 2024 he suspended active campaigning and endorsed Trump, explicitly acknowledging he had no realistic path to victory; at that time public polling put him around 2%.(en.wikipedia.org) His final vote share under 1% confirms that his effort did not evolve into a major, system‑challenging force.
• Overall third‑party impact: Post‑election analyses and official tallies consistently characterize third‑party and independent performances as minor—third‑party totals remained below 5% nationally and did not alter the two‑party lock on the Electoral College.(en.wikipedia.org) The 2024 results look like a conventional two‑party race with only small third‑party leakage, not a year in which an outsider effort “challenged the traditional two‑party split.”