Big Beautiful Bill, Elon/Trump, Dollar Down Big, Harvard's Money Problems, Figma IPO

Multiple contemporaneous reports and legislative records show that the “Big Beautiful Bill” (formally the One Big Beautiful Bill Act) passed the House in its final reconciled form on July 3, 2025, and was signed into law by President Donald Trump on July 4, 2025.(en.wikipedia.org) This matches Jason’s prediction that the bill would pass the House and be signed into law by Trump by July 4, 2025, so the prediction was correct.

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).

The prediction is explicitly about what will happen “over the next several years”, but we are less than half a year past the July 4, 2025 podcast date, so the forecast horizon has not elapsed.

Regulatory structure is still unsettled:

• States are rapidly enacting their own AI laws: in 2025, state lawmakers introduced 1,134 AI‑related bills and enacted 131 laws across 40 states, creating a patchwork of state rules. (ccianet.org)
• At the same time, there is an active federal push to preempt state AI laws (e.g., House bills including 10‑year preemption, and a White House draft executive order to challenge state AI regulations), while state AGs and groups like NCSL are lobbying hard against preemption. (nfcommission.gov)
  Because this federal–state fight is unresolved, we don’t yet know whether the medium‑term U.S. regime will truly be “predominantly state‑level, non‑preempted” as the prediction assumes.

Impact on startups vs incumbents is also not yet empirically clear:

• California’s SB 53 (the first major “frontier AI” safety law) and related AI bills are widely seen as positioning California as a lead U.S. regulator of AI, but SB 53 is narrowly scoped: it applies only to “large frontier developers” with models above 10^26 FLOPs and >$500M in annual revenue, deliberately excluding smaller developers. (lw.com)
• Critics from venture capital and industry (e.g., Andreessen Horowitz) argue that a 50‑state patchwork could make compliance costs unmanageable for startups and entrench big players, explicitly warning that SB 53 and similar state‑by‑state rules may “squeeze out startups” and “entrench the biggest players.” (forbes.com)
• However, supporters of SB 53 and academic/policy analysis emphasize that its obligations fall almost entirely on a small set of frontier labs and are not designed to burden smaller startups, and there is not yet systematic evidence that AI startup formation, funding, or growth has materially slowed because of these new state laws. (sb53.info)

Since:

1. the regulatory regime Chamath is describing is still being contested (federal preemption vs state autonomy), and
2. the concrete, measurable long‑term impact on AI startups and incumbents has not yet had time to materialize or be robustly measured,

it is too early to say whether his multi‑year prediction is right or wrong. Hence the result is inconclusive at this time.

As of November 30, 2025, it is not yet 2032–2033, so we cannot know whether new U.S. nuclear power plants whose development is initiated around 2025 will, in fact, start operating before those years. The prediction concerns events (plant commissioning dates) that are still in the future relative to today, so it cannot be verified or falsified yet. Therefore, the correct status is too early to tell.

The prediction is explicitly about a ~20‑year horizon: from the “next 20 years” after the mid‑2020s (i.e., roughly through 2045). As of November 30, 2025, less than one year has passed since the podcast release date (July 4, 2025), leaving about 19 more years until the prediction’s time window is complete. Because the claim concerns long‑term investment performance across that entire period, it cannot yet be evaluated—neither the full trajectory of U.S. electricity production investments nor the long‑run trend of marginal energy costs toward zero is observable for 2045. Therefore, the correct status today is that it’s too early to tell whether the prediction is right or wrong.

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.

Available reporting shows that after a severe mid‑2025 rupture over Trump’s “One Big Beautiful Bill,” Trump and Musk ultimately moved back toward cooperation rather than remaining long‑term adversaries.

• In June–July 2025, Musk broke dramatically with Trump over the “Big, Beautiful Bill,” denouncing it as “insane” deficit spending, threatening primary challenges and creating a new “America Party” in explicit rebuke to Trump and the GOP. Trump in turn called Musk “off the rails” and a “train wreck,” and threatened to cut subsidies or even look at deporting him, with coverage characterizing the split as a spectacular falling‑out. (livemint.com) This period clearly looked like a real rupture.
• However, the America Party effort quickly stalled; by late July reporting noted Musk hadn’t filed FEC paperwork or fielded candidates, and by August 20 he was already “giving up” on actually starting the party and instead considering support for VP JD Vance in 2028—signaling retreat from a sustained, organized challenge to Trumpism rather than a permanent realignment against it. (en.wikipedia.org)
• On September 21, 2025, Trump and Musk publicly met, shook hands and chatted at a large memorial for Charlie Kirk in Arizona. The Guardian and other outlets framed this as raising the prospect of reconciliation, and noted that Musk himself posted a friendly image of the two together captioned “For Charlie,” while the White House account amplified the encounter. (theguardian.com) The Wikipedia entry on the Trump–Musk feud summarizes the second phase of their conflict (June 28–September 21, 2025) as ending with a “Reunited” status, reflecting that the feud, as such, was considered over. (en.wikipedia.org)
• By late 2025, Musk was again doing business with the administration—e.g., a GSA agreement with his xAI for federal AI adoption—indicating a pragmatic working relationship despite unresolved policy differences over the bill. (en.wikipedia.org)

Taken together, the sequence is: intense but time‑bounded feud, an attempted but quickly deflated third‑party break, then an early‑fall public rapprochement and resumed cooperation. That matches Chamath’s directional claim that their conflict would be temporary and that they would ultimately find common ground rather than experience a lasting rupture, even though the path there involved a much deeper and more chaotic break than his framing implied. Hence the prediction is best scored as right in outcome, albeit messier and more contingent than suggested in the podcast.

The prediction is explicitly framed over “the coming decades” and most of the current generation’s lifetimes, so only a few months of data (July–November 2025) are far too short to validate or falsify it.

So far, events are directionally consistent with Chamath’s thesis but cannot prove it:

• The U.S. dollar has experienced a historically sharp drop in 2025 (around 10–11% on the dollar index in the first half of the year), but analysts and strategists mostly describe this as a significant devaluation and period of stress, not a structural collapse; many still expect a gradual grind lower rather than a sudden failure, and central banks continue to treat the dollar as the primary reserve asset. (theinvestorschronicle.com)
• U.S. equities (e.g., the S&P 500) are up roughly 12–15% year‑to‑date in 2025, indicating that major dollar‑denominated assets have indeed appreciated faster than the dollar’s recent decline so far. (statmuse.com)

However, because the claim concerns multi‑decade currency behavior and long‑term real returns on U.S. assets, the available post‑podcast window (less than one year) is nowhere near sufficient to declare it right or wrong. The correct assessment today is that it’s too early to tell.

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).