Biden chaos, Soft landing secured? AI sentiment turns bearish, French elections

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.

Evidence from mid‑2024 through mid‑2025 shows exactly the kind of turbulent, sorting-out period Chamath described:

• Volatile and selective capital markets for AI startups:

  • In Q2 2024, funding for Indian AI startups collapsed 91% quarter‑over‑quarter and 82% year‑on‑year, even as US AI funding surged – a clear sign of instability and regional selectivity right as his predicted window began. (economictimes.indiatimes.com)
  • Globally, AI appeared “hot” in headline dollars, but funding became highly concentrated in a few mega‑rounds (OpenAI, Anthropic, Databricks, xAI), while overall VC activity remained far below 2021 and exits were sparse, indicating a difficult backdrop despite the top‑line boom. (barrons.com)
  • By 2025, PitchBook data shows the number of AI‑related deals falling to a five‑year low even as AI takes a record share of VC dollars, with investors “prioritizing scale over diversity” and warning of high failure rates and an exit crunch—classic signs of a shakeout rather than an easy market. (ainvest.com)
  • Major investors and analysts openly warned of an AI valuation bubble and overinvestment risk in early 2025, underscoring that capital markets were nervous and uneven, not smooth. (reuters.com)

• Visible shakeout of weak or overhyped AI companies within ~1 year of his July 2024 call:

  • Multiple high‑profile, well‑funded AI or AI‑branded startups either shut down, entered insolvency, or were forced into fire‑sale style outcomes in late 2024 and 2025: Ghost Autonomy (autonomous driving, shut April 2024), Humane (AI Pin hardware startup, effectively wound down and acquired by HP after a failed product), Builder.ai (once a unicorn, in insolvency proceedings and 80% layoffs by mid‑2025), and Sunshine (Marissa Mayer’s consumer AI startup, shut down with assets sold to a new vehicle). (aimresearch.co)
  • Regional reports document smaller AI startups in India and elsewhere simply running out of runway or pivoting away from their original AI theses (e.g., Subtl.ai shutdown; InsurStaq.ai closing after a year; Unikon.ai burning cash, failing to raise follow‑on, and shutting down). These are precisely the “overfunded or weak” players being exposed. (theoutpost.ai)
  • Even strong, well‑capitalized AI infrastructure companies like Scale AI had to lay off staff and sharply restructure their generative‑AI efforts in 2025, citing overexpansion and “market shifts,” another sign that easy money and unchecked headcount growth were no longer being rewarded. (businessinsider.com)
  • A widely cited market analysis in late 2025 explicitly describes a deflating AI startup bubble and a selective correction, with Tier‑1 hyperscalers and a small set of unicorns surviving, while a broad Tier‑3 of AI startups is failing or struggling—very close to Chamath’s “reckoning” and “sorting out who’s actually real.” (marketwatch.com)

Putting this together, from roughly Q3 2024 through at least mid‑2025 there were multiple “complicated quarters”: funding highly uneven and concentrated, deal counts and exits weak, and a growing list of shutdowns, insolvencies, and capitulations among AI startups. That pattern matches Chamath’s forecast of 2–3 tough quarters and about a year of turbulence and shakeout in AI‑related startup and capital markets, so the prediction is best judged as right.

Apple did release the iPhone 16 in September 2024 with its broader Apple Intelligence branding, but the core scenario Sacks described never materialized:

• The advanced Siri overhaul based on large‑language‑model tech that was advertised alongside iPhone 16 was delayed to 2026, not shipped with iPhone 16. Apple executives and reporting (e.g., Bloomberg, summarized by MacRumors and others) note that the personalized, LLM‑driven Siri features slipped to an iOS 26.4 release in spring 2026 after quality problems, despite being used in iPhone 16 marketing. (macrumors.com)
• Instead of relying solely on an Apple‑built on‑device LLM, Apple is now planning to power major parts of revamped Siri with Google’s Gemini models running in Apple’s Private Cloud Compute, handling planning and summarization for Siri with a target around March 2026. That means the Siri upgrade is not purely on‑device nor purely Apple’s own model, contrary to the prediction’s premise about "rolling their own LLM" and not sharing data with third‑party LLMs. (androidcentral.com)
• On the demand side, analyst work from UBS finds that last year’s iPhone 16 and its Apple Intelligence branding did not spur a widespread upgrade cycle: the average iPhone in use (outside China) remained about 37 months old, and UBS explicitly concluded there is “no groundbreaking AI use case” and no significant AI‑driven sales boost, contrary to expectations of a massive upgrade wave. (barrons.com)
• Broader coverage of Apple’s 2025 software cycle similarly describes investors as underwhelmed by Apple’s AI progress and skeptical that current AI features will drive a major iPhone upgrade boom; Apple’s market cap fell on news of Siri delays, and analysts characterized updates as incremental rather than super‑cycle‑inducing. (businessinsider.com)

Given that (1) the exact condition Sacks specified—an iPhone 16 with a working, Apple‑only on‑device LLM powering Siri—still hasn’t occurred, and (2) the actual iPhone 16 + Apple Intelligence launch has not led to “everyone” upgrading or a clearly massive AI‑driven upgrade cycle, the prediction about a huge iPhone 16 AI upgrade wave is best judged wrong based on the real‑world outcome to date.

The quote and context check out: in the July 12, 2024 All‑In episode, Sacks says Glue’s LLM‑powered results are “kinda like a B plus right now” and adds that “in a year or so… the next gen version of all the models… At that point, it’s going to be an A plus.”(podscripts.co)

Within roughly a year, his “next gen” condition did occur. OpenAI released several successor models to GPT‑4, including GPT‑4.5 in February 2025 and GPT‑4.1 in April 2025, followed by GPT‑5 and GPT‑5.1 later in 2025; these were marketed and benchmarked as substantially more capable and reliable than GPT‑4/4o, with improved coding, instruction following, and lower hallucination rates on certain tests.(en.wikipedia.org)

However, whether that constitutes "A+" practical quality for Glue‑style applications is inherently subjective and not directly measurable from public data:

• No objective Glue metrics – Public coverage of Glue describes its AI as a virtual employee layered on top of Slack/Teams‑like chat, powered by models such as ChatGPT and Claude, but there are no later public benchmarks or statements from Sacks quantifying that its outputs have become “A+” versus the earlier “B+.”(techcrunch.com)
• Hallucinations and reliability remain open issues – Even by late 2024 and through 2025, peer‑reviewed work and industry analysis still describe hallucinations as a “critical barrier” for enterprise use, with state‑of‑the‑art methods often failing to exceed ~80% factual faithfulness on some benchmarks.(arxiv.org) Studies and reporting in 2025 note that newer reasoning models like OpenAI’s o3 and o4‑mini can hallucinate in 30–50% of cases on certain tasks, and major outlets emphasize that leading LLMs still generate incorrect or fabricated content often enough to demand human verification.(livescience.com)
• Mixed real‑world sentiment – Aggregated user‑review data shows that mentions of hallucinations in reviews of leading chatbots (ChatGPT, Gemini, Claude, etc.) dropped markedly between March and October 2025, but did not disappear; non‑trivial shares of users still report accuracy and hallucination concerns.(learn.g2.com)

Because “B+ vs A+” is Sacks’s own informal grading with no standardized threshold, and there is no transparent data on how Glue’s actual output quality changed relative to that scale, yet frontier models clearly did improve while still exhibiting meaningful errors, the prediction cannot be judged definitively true or false from available evidence. That fits best under “ambiguous” rather than “right” or “wrong.”

Jason’s prediction was that by sometime in 2025 AI tools would hit a tipping point in adoption such that many organizations achieve substantial labor arbitrage (one worker effectively doing the work of several / replacing large chunks of lower‑skilled knowledge work).

What we do see by late 2025

• Multiple surveys show very broad individual adoption of AI at work. Microsoft’s Work Trend Index data, summarized by Worklytics, reports that about 75% of global knowledge workers are now using AI tools regularly, with usage roughly doubling in the prior six months. (worklytics.co) A Clutch survey finds 74% of full‑time workers regularly use AI on the job. (lifewire.com)
• Microsoft’s 2025 Work Trend Index for Switzerland reports that 52% of Swiss organizations already use AI agents to automate business processes, and 72% of leaders plan to deploy AI agents as “digital team members” to expand workforce capacity over the next 12–18 months—clear signs of serious organizational adoption and intent to use AI as a capacity multiplier. (news.microsoft.com)
• McKinsey’s 2025 "AI in the workplace" report finds that 94% of employees and 99% of C‑suite leaders have some familiarity with gen‑AI tools, and about 13% of employees already use gen‑AI for more than 30% of their daily work. The report explicitly describes the situation as “beyond the tipping point” in terms of awareness and early use. (mckinsey.com)
• Field and survey studies show real productivity effects: a 6‑month randomized experiment across ~6,000 knowledge workers found that those with an integrated gen‑AI assistant spent about 25% less time on email and finished documents somewhat faster, demonstrating that some employees can handle more output in the same hours. (arxiv.org) A systematic review of practitioner studies reports that many professionals delegate routine tasks to gen‑AI and sometimes bypass peers/subordinates in favor of AI, indicating nascent task‑level substitution. (arxiv.org)
• Executives increasingly talk in explicitly labor‑arbitrage terms. A BearingPoint survey reports that around half of executives believe their firms are already 10–19% overstaffed due to automation and AI and expect overcapacity could reach up to 50% in three years, especially in back‑office, customer service, and entry‑level finance/HR roles. (techradar.com) This shows that some organizations perceive AI as enabling fewer people to cover the same workload.

What we do not clearly see yet

• The same McKinsey 2025 report emphasizes that, despite high familiarity, most organizations are still moving slowly from pilots to scaled deployment; 47% of C‑suite leaders say their organizations are releasing gen‑AI tools too slowly and are still building or refining roadmaps. (mckinsey.com) That is more consistent with early transformation than a completed tipping point in structural labor arbitrage.
• Research on job exposure finds that nearly all jobs have some generative‑AI exposure but only a minority are heavily affected so far. The GAISI study for the UK shows widespread exposure but relatively limited heavily affected roles, and early evidence that displacement effects may be emerging, not yet dominant. (arxiv.org)
• Macro‑level analyses from McKinsey estimate that half of current work activities could be automated only between about 2030 and 2060, even under favorable adoption scenarios, implying that the full labor‑substitution effects unfold over decades rather than having clearly arrived by 2025. (mckinsey.com)
• Several worker surveys report that AI often adds oversight and correction work and can increase workload, rather than simply letting one person replace multiple others outright, at least with current tools and governance. (news.com.au)

Why the outcome is ambiguous rather than clearly right or wrong

• On the “right” side: by 2025, AI is indeed widely deployed; many knowledge workers use it daily; some firms are automating entire processes; some executives explicitly see staff overcapacity from AI; and individual‑level studies show meaningful productivity gains that plausibly allow fewer staff to cover the same work in certain functions. These facts support Jason’s intuition that AI would start enabling labor arbitrage by 2025.
• On the “wrong” side: the prediction implied a broad, clear tipping point where “a lot” of labor arbitrage occurs across many organizations, with one worker routinely doing the work of several or large portions of lower‑skilled knowledge work being systematically replaced. Available evidence shows early and uneven adoption, modest but real task substitution, and mostly hybrid "human + AI" workflows, not a well‑documented, economy‑wide flip to large‑scale labor replacement in 2025.
• Crucially, terms like “tipping point,” “a lot of labor arbitrage,” and “many organizations” lack precise thresholds, and existing data does not cleanly map onto them. Different reasonable observers could look at the same mix of high adoption, early job restructuring, but still‑limited measured displacement and disagree on whether Jason’s bar has been met.

Because substantial evidence supports partial realization of his claim (especially in some sectors and early‑adopting firms), but there is no clear, quantitative demonstration that a broad 2025 tipping point in labor arbitrage has occurred, the prediction cannot be judged cleanly true or false. Hence the evaluation: ambiguous.

Since the July 12, 2024 podcast, several humanoid and advanced general-purpose robots have indeed reached roughly the US$20,000 price point and become commercially offered, matching Jason’s order‑of‑magnitude prediction.

Concrete examples:

• 1X Technologies’ NEO home humanoid robot opened global preorders in October 2025 at a listed price of US$20,000 (or a $499/month subscription), with deliveries scheduled for 2026, marketed explicitly as a consumer-ready humanoid home assistant.​​ (en.wikipedia.org)
• Chinese firms are also targeting similar or lower prices: Unitree’s R1 humanoid was announced in 2025 at about US$5,900, and its other humanoids (G1, H1) plus competing platforms like Pudu’s D9 and Neura’s 4NE‑1 cluster in the mid‑five‑figure or sub‑US$20k–30k band, placing advanced humanoids in the same order of magnitude as US$20,000.​ (nypost.com)
• UBTech publicly discussed selling a home/elder‑care humanoid at around US$20,000, further corroborating that manufacturers see ~US$20k as a realistic commercial price point for such robots in this timeframe.​ (cybernews.com)
• While Tesla Optimus itself is not yet sold to outside customers, Elon Musk and Tesla have repeatedly framed Optimus’s intended price in the roughly US$20,000–30,000 range, reinforcing that this is the expected commercial band for Optimus‑class humanoids.​ (digitimes.com)

Because humanoid and Optimus‑class robots are now actually being offered for sale or preorder at ~US$20,000 in the 2024–2026 window that would reasonably count as the "foreseeable near term" from the podcast date, the substance of Jason’s prediction has come true.

Evidence from July 2024 through mid‑2025 supports Chamath’s core claim that the post‑election National Assembly was so fragmented that neither Marine Le Pen’s far‑right National Rally (RN) nor Jean‑Luc Mélenchon’s New Popular Front (NFP) could push through their major reform agendas, and that France endured extended political logjam.

Key points:

• The June–July 2024 snap legislative election produced a hung parliament split into three hostile blocs (NFP, Macron’s centrist Ensemble, and RN), with none close to a majority. This fragmentation led to a prolonged political crisis and a succession of weak minority governments.(en.wikipedia.org)

• Michel Barnier’s government fell in December 2024 after using Article 49.3 on a social‑financing bill; parliament then passed an emergency law simply rolling over the previous year’s budget with no policy changes, underscoring the inability to agree on substantive legislation.(en.wikipedia.org)

• François Bayrou’s successor government only managed to get a 2025 budget in place in early 2025, relying on constitutional shortcuts and a tenuous non‑aggression pact. That budget focused on deficit reduction via large spending cuts and tax increases, not on implementing RN’s hard‑line immigration program or the NFP’s expansive left‑wing economic platform (retirement at 60, big minimum‑wage hike, price freezes, wealth‑tax overhaul, etc.), which remained largely aspirational.(reuters.com)

• Throughout 2024–25, international and French press repeatedly described France as being in political deadlock, with a parliament that is “hard to govern” and governments that struggle or fail to secure majorities for their agendas, especially on reforms beyond the bare‑minimum budgetary measures demanded by fiscal pressures.(euronews.com)

• Only in late 2025—more than a year after the elections—did the Assembly manage a notable policy shift on a headline issue (temporarily suspending the 2023 pension reform), and even that fell short of the NFP’s goal of full repeal and a new retirement age of 60.(apnews.com)

Taken together, this shows: (1) the Assembly was indeed highly fragmented; (2) neither Le Pen’s nor Mélenchon’s flagship reform programs were enacted; and (3) for roughly the first year after the election, lawmaking was dominated by crisis management and stop‑gap or fiscally driven budget measures rather than transformative reforms. While one can argue that passing an austerity budget is itself a significant policy move, the spirit of Chamath’s prediction—that France would lose about a year of real reform to political logjam—matches what actually happened, so the prediction is best judged as right overall.

Available reporting around July 12–22, 2024 shows intense media and political scrutiny of Biden’s health, but no named whistleblower from Biden’s own inner circle publicly coming forward with evidence of a cover‑up within that 10‑day window.

Key facts:

• On July 8, 2024 (before the podcast window even starts), The Wall Street Journal reported that Biden’s team had long limited his schedule and unscripted interactions to minimize concerns about age and mental acuity, based on interviews with about 50 people. These were largely anonymous sources and the story predated the July 12 release date. (en.wikipedia.org)
• In early July 2024, multiple outlets reported that a Parkinson’s specialist repeatedly visited the White House, based on visitor logs; the White House physician, Dr. Kevin O’Connor, publicly denied that Biden was being treated for Parkinson’s and insisted examinations found no such disorder. This was document‑ and log‑driven reporting, not a Biden‑inner‑circle whistleblower alleging a cover‑up. (theweek.com)
• On July 10, 2024, House Oversight subpoenaed three close aides (Annie Tomasini, Anthony Bernal, and Dr. O’Connor) after an unnamed former Biden aide alleged they were “seeking to cover up” Biden’s declining cognitive state. That allegation was relayed by Republicans and media, but the source remained anonymous and there is no record of a public, on‑the‑record whistleblower from Biden’s inner circle in the following 10 days. (dailywire.com)
• Biden announced he would not seek re‑election on July 21, 2024, but mainstream acknowledgment that his “inner circle worked to conceal his decline” came later via a Washington Post editorial on August 20, 2024, and still was not tied to a single inner‑circle whistleblower stepping forward at that time. (en.wikipedia.org)
• More detailed claims of a cover‑up based on interviews with inner‑circle figures emerged only in 2025 with books like Original Sin (Tapper & Thompson), drawing on roughly 200 mostly anonymous interviews conducted after the 2024 election. (en.wikipedia.org)

Because no named, public whistleblower from Biden’s medical staff, close aides, or senior White House personnel came forward between July 12 and about July 22, 2024 with evidence alleging a cover‑up of his cognitive decline, Jason’s time‑bound prediction **did not come true.

Chamath’s prediction was that before the 2024 U.S. presidential election there would be a “Deep Throat / Watergate‑style” investigative leak by a well‑placed insider working with major journalists, exposing a deliberate cover‑up of President Biden’s cognitive condition.

What actually happened:

1. Prior to and during the 2024 campaign, there was no singular, Woodward‑and‑Bernstein‑style exposé that definitively “got to the bottom” of an organized cover‑up of Biden’s cognition. The most detailed accounts of an alleged systematic cover‑up came later, in 2025 books such as Original Sin: President Biden’s Decline, Its Cover-Up, and His Disastrous Choice to Run Again by Jake Tapper and Alex Thompson, and Chris Whipple’s Uncharted, which rely on extensive retrospective sourcing and describe efforts inside Biden’s circle to conceal or minimize his decline. These were published months after the November 5, 2024 election. (en.wikipedia.org)

2. The closest analogue—Carl Bernstein’s July 1–2, 2024 reporting—predated the podcast and fell short of the predicted “Watergate moment.” In early July 2024, Bernstein reported on CNN (and was widely summarized in outlets like the New York Post and others) that multiple sources close to Biden said there had been 15–20 episodes over the prior 18 months where Biden appeared as impaired as he did in the June 27 debate, and that concerns had been raised with former chief of staff Ron Klain and “pushed back.” (nypost.com) While significant, this was:

   • based on anonymous, second‑hand sources, not a named Deep‑Throat‑style whistleblower; and
   • a single TV segment and follow‑up write‑ups, not a sustained investigative series uncovering the full machinery of an official cover‑up. Importantly, this reporting aired before the July 12, 2024 podcast episode, so it cannot be the future event Chamath was forecasting.

3. The Robert Hur special‑counsel report (February 2024) raised cognitive concerns but was an official DOJ product, not a journalistic leak. Hur described Biden as a “well‑meaning, elderly man with a poor memory” and recounted notable memory lapses, prompting intense political debate about Biden’s mental fitness. (theguardian.com) However, this was a formal prosecutorial report, not a clandestine leak from an insider to investigative reporters exposing a hidden cover‑up campaign.

4. Post‑election retrospectives explicitly frame the big revelations as coming too late, not as something voters fully learned before November 2024. In promoting Original Sin in 2025, Jake Tapper has repeatedly said that legacy media—including himself—under‑covered Biden’s decline in real time and is now trying to reckon with that failure. (realclearpolitics.com) That admission is difficult to square with the idea that there had already been a decisive Watergate‑like media exposé before the 2024 vote.

Given these facts, there was no single, decisive “Deep Throat / Watergate‑style” investigative leak before the 2024 election that clearly exposed a coordinated cover‑up of Biden’s cognitive condition in the way Chamath described. Instead, the more thorough accounts of any alleged cover‑up emerged in 2025 books and subsequent congressional probes, well after the election had been decided.

Therefore, the prediction is wrong.

Key parts of Jason’s prediction both did and did not line up with what actually happened, and one of his central claims can’t be directly tested.

1. Biden did not remain the Democratic nominee, so “Trump will beat Biden” never occurred as stated.
Joe Biden ran for re‑election, won the Democratic primaries, and secured enough delegates to be the nominee, but on July 21, 2024 he withdrew from the race and endorsed Kamala Harris.(en.wikipedia.org) Because Biden stepped aside before the general election, there was no Biden‑vs‑Trump contest; Trump instead ran against Harris and defeated her.(en.wikipedia.org) That means the specific forecast “Trump will beat Biden in 2024” cannot be directly verified or falsified.

2. The “speed run” / rapid replacement process essentially did happen, but not as a voter‑facing primary.
After Biden’s withdrawal and endorsement, Democrats coalesced around Harris with remarkable speed:

• Within about a day, Harris had enough delegate commitments to clinch the nomination.(people.com)
• The DNC then set up an accelerated virtual roll‑call process (Aug 1–5) to officially nominate her before the in‑person August convention, explicitly to meet timing and ballot‑access concerns and avoid a messy open convention.(theguardian.com)
• Groups like Black Lives Matter criticized this as an elite‑driven, rushed process that bypassed broader voter participation, and even demanded a more open virtual primary—underscoring how compressed and top‑down the replacement process was.(reuters.com)

So, Jason’s broad idea that Democrats would respond to Biden’s weakness with a fast, improvised nomination “speed run” was largely borne out, though it took the form of a rapid delegate realignment and virtual roll‑call rather than a traditional, multi‑candidate primary.

3. Overall assessment

• The election‑outcome claim (“Trump will beat Biden”) hinges on a matchup that never occurred because Biden withdrew; that part is not verifiable.
• The process‑prediction (a rapid, improvised effort to replace or re‑legitimize the nominee) broadly matches what happened with Harris and the DNC’s compressed virtual nomination.

Because part of the prediction is untestable and another part is substantially but not perfectly accurate, the fairest judgment is “ambiguous” rather than clearly right or wrong.

The prediction has two parts, both of which must occur:

1. A fast, highly compressed Democratic process to choose/ratify a new nominee by Aug. 19, 2024
   This did happen. After Joe Biden withdrew on July 21, 2024, Democrats rapidly organized an accelerated selection process using a virtual roll call:

   • On July 22 the DNC confirmed it would use a pre‑convention virtual roll call to select the nominee and have one in place by Aug. 7, explicitly emphasizing that the process would be “fast.” (axios.com)
   • On July 30, the DNC formally announced that only Kamala Harris qualified and that the virtual roll call to nominate her would run Aug. 1–5. (democrats.org)
   • Harris was officially certified as the Democratic presidential nominee on Aug. 5, 2024, before the in‑person convention opened on Aug. 19. (en.wikipedia.org)
     This fits the “speedrun” / highly compressed selection process part of the prediction.

2. A whistleblower about Biden’s cognitive state coming forward publicly in July 2024
   Extensive July 2024 coverage documented concerns about Biden’s age and cognition, based on reporting and anonymous current/former aides, but there was no widely reported, on‑the‑record whistleblower in that month:

   • Summaries of July 2024 events list media investigations (e.g., Wall Street Journal and Washington Post pieces) and public calls from Democrats for Biden to step aside, but they do not identify a specific named whistleblower who publicly came forward about a cognitive cover‑up in July. (en.wikipedia.org)
   • Key insiders such as White House physician Kevin O’Connor and Jill Biden adviser Anthony Bernal were drawn into congressional inquiries via letters and subpoenas (e.g., July 7 and July 10, 2024), but they did not themselves step forward as whistleblowers in that period. (en.wikipedia.org)
   • Whistleblower‑style accounts about Biden’s mental decline from Secret Service or other insiders surface later (in 2025 reporting and book coverage), not in July 2024. (thedailybeast.com)

Because the “speedrun” nomination process prediction was essentially right but the July 2024 whistleblower part did not occur, the overall conjunctive prediction is wrong.

Joe Biden did in fact withdraw from the 2024 race on July 21, 2024, giving up his status as the presumptive Democratic nominee and endorsing Kamala Harris shortly afterward. (en.wikipedia.org)

After Biden’s exit, Democrats did not run a new, open, voter-facing primary or “speedrun” primary. Instead, the DNC set up a delegate petition process for the nomination: candidates needed at least 300 delegate signatures to qualify for the ballot. Only Kamala Harris met that threshold, with 3,923 delegates backing her; no other candidate qualified. (democrats.org) Delegates then participated in a virtual roll call from August 1–5, 2024, in which Harris received roughly 99% of votes and became the official Democratic nominee. (cbsnews.com)

Critics and later retrospectives explicitly described the switch as replacing Biden with Harris without holding an open primary and lamented that there hadn’t been one, underscoring that no genuine open or rushed primary contest ever occurred once Biden stepped aside. (en.wikipedia.org)

Given that (1) Biden stepped down, (2) Democrats did not hold an open/speedrun primary, and (3) Kamala Harris was selected and became the nominee through a largely uncontested delegate process, Chamath’s conditional prediction matches what actually happened.