E132: SEC goes after crypto giants, Sequoia splits, LIV/PGA, Messi's deal + LIVE Q&A!

Key parts of Chamath’s prediction did not come true, even though some short‑term elements were directionally right.

1. ETF freeze claim clearly failed

• On January 10, 2024, the SEC approved 11 spot bitcoin ETFs from major issuers like BlackRock, Fidelity, and others, allowing U.S. investors direct exposure to spot bitcoin via ETFs. (apnews.com)
• On May 23, 2024, the SEC approved rule changes so exchanges could list multiple spot ether ETFs, further expanding crypto ETF access. (cnbc.com)
• By 2025, the SEC had also approved in‑kind creation/redemption mechanisms for all spot bitcoin and ether ETFs, making them even more institutional‑friendly. (coindesk.com)

Chamath predicted the SEC “will not approve any ETFs” during the crackdown; in reality, the agency approved major spot crypto ETFs less than a year after his June 2023 statement. That’s a direct contradiction.

2. “Systematic dismantling” of the whole value chain did not fully happen

• Exchanges & staking: The SEC did intensify enforcement against big exchanges and staking offerings in 2023–24: lawsuits against Coinbase and Binance for operating as unregistered exchanges and offering unregistered staking programs, as well as actions involving Genesis/Gemini, Celsius, Kraken, Nexo, and later Consensys’ MetaMask Staking. (en.wikipedia.org) This matches the early part of his value‑chain narrative (exchanges and staking services).
• Custodial services: The SEC pushed on custody mainly via rules and advisers rather than “systematically dismantling” custodians. It proposed a stricter custody rule in 2023 and brought a 2024 case against Galois Capital for failing to keep fund crypto with a qualified custodian, but by late 2025 it issued a no‑action letter easing the ability of state trust companies to act as crypto custodians. (cnbc.com) This is a mix of pressure and later accommodation, not a linear crackdown down the custody stack.
• Venture firms / staking‑based token deals: There is no evidence the SEC followed through with a targeted campaign against venture funds for their token‑staking schemes or founding‑coin allocations. Instead, venture players like a16z’s crypto arm show up as stakeholders in policy discussions and industry lawsuits challenging the SEC, not as primary enforcement targets over staking economics. (reuters.com)
• Reversal of the crackdown: Under the Trump administration (from early 2025), the SEC created a crypto task force and then dismissed or paused major lawsuits against Coinbase, Binance, Kraken, and others, and issued guidance softening its stance on many staking models. (reuters.com) That is the opposite of a continuing “systematic dismantling” through the entire value chain.

Because (a) the SEC did approve major crypto ETFs and (b) the enforcement campaign did not proceed all the way through custodians and the venture ecosystem—and was partly rolled back in 2025—the overall prediction is best judged wrong, despite being partially accurate about an initial enforcement surge against exchanges and staking services.

The bill Jason was referring to is H.R. 2797, the Equal Opportunity for All Investors Act of 2023, which would let individuals qualify as accredited investors by passing an SEC‑designed exam. It passed the U.S. House on May 31, 2023 and was received in the Senate on June 1, 2023, but no further action was taken and the bill died at the end of the 118th Congress without a Senate vote or presidential signature. (congress.gov) A substantially similar bill, H.R. 3339 (Equal Opportunity for All Investors Act of 2025), passed the House on July 21, 2025 and was referred to the Senate on July 22, 2025, where it still has only “Passed House” status and has not become law as of November 30, 2025. (congress.gov) Coverage of H.R. 3339 explicitly notes that the 2023 version died in the Senate, confirming that Jason’s expectation that the 2023 House‑passed bill would soon pass the Senate and be signed into law did not occur within the next legislative session or by the current date. (them.us)

Jason predicted that, due to U.S. regulatory pressure, Coinbase CEO Brian Armstrong would move Coinbase’s corporate base of operations out of the United States, likely to the Middle East or a Caribbean island.

What actually happened:

1. Armstrong explicitly said Coinbase would not relocate overseas. In May 2023 he publicly rowed back earlier hints and stated that “Coinbase is not going to relocate overseas” and that the company would always maintain a U.S. presence, emphasizing long‑term commitment to the U.S. market. (cnbc.com)

2. Coinbase remains a U.S. company and has not shifted its corporate base abroad. Coinbase Global, Inc. continues to be a U.S.-based, publicly listed company; its primary corporate identity and operations remain anchored in the United States. (en.wikipedia.org)

3. The main legal move was from one U.S. state to another, not overseas. In November 2025, Coinbase announced it would move its legal incorporation and headquarters structure from Delaware to Texas, citing Texas’s more favorable corporate and regulatory environment. This is an intra‑U.S. move, not a relocation to the Middle East or Caribbean. (reuters.com)

4. International expansion did occur, but only via subsidiaries, not by moving the corporate base. Coinbase launched Coinbase International Exchange in Bermuda and built it out as an offshore derivatives platform for non‑U.S. clients, describing Bermuda as one of its international hubs. This is framed as global expansion while remaining “committed to the U.S.”, not as a shift of the parent company’s headquarters or corporate domicile. (coinbase.com)

5. No evidence of a Middle East or Caribbean HQ move. While Armstrong has praised jurisdictions like the UAE and said Coinbase is interested in investing more internationally, he has simultaneously affirmed that leaving the U.S. is “not even in the realm of possibility” and that there is no contingency plan to relocate the company out of the U.S. (cointeeth.com)

Given that, as of November 30, 2025, Coinbase has not moved its corporate base of operations out of the United States—let alone to the Middle East or a Caribbean island—the prediction did not come true.

Chamath argued in June 2023 that there was not enough political support to "visit this issue" and that he was "pretty skeptical that you're going to see any form of legislation pass" on a clear federal crypto framework.

Since then, Congress has in fact passed significant federal crypto legislation. In 2025, the Guiding and Establishing National Innovation for U.S. Stablecoins (GENIUS) Act was passed by both chambers (Senate on June 17, 2025; House on July 17, 2025) and signed into law by President Trump on July 18, 2025. (en.wikipedia.org) The law is explicitly framed as a federal regulatory regime for payment stablecoins—establishing licensing, reserve, disclosure and supervisory requirements—and is widely described by legal and policy analysts as the first major federal law focused on cryptocurrency regulation and a clear framework for stablecoins. (mondaq.com) Government and industry commentary similarly treat it as a landmark crypto regulatory framework rather than a marginal or symbolic bill. (ffnews.com)

While a broader market‑structure bill (the Financial Innovation and Technology for the 21st Century Act, FIT21) only passed the House and never became law, (en.wikipedia.org) the GENIUS Act itself is clearly significant, sector‑defining crypto legislation passed by Congress within roughly two years of his prediction. That contradicts his expressed skepticism that there was enough political support for any such legislation to pass.

Because a major federal crypto framework law did pass in this timeframe, the prediction that no such legislation would pass is best classified as wrong, even allowing for some ambiguity around the phrase "near term."

The prediction is explicitly multi‑decade: Chamath says China is “largely uninvestable for the next 30 or 40 years,” which from a June 2023 starting point implies a time horizon stretching roughly into the 2050s–2060s. As of November 30, 2025, only about two and a half years of that 30–40 year window have elapsed, so the core claim—that over that entire span China will deliver materially worse long‑term, risk‑adjusted equity returns than major alternative markets—cannot yet be evaluated.

Early performance differences or current investor sentiment toward China (positive or negative) are not sufficient to confirm or refute a structural, multi‑decade return forecast. We would need many more years of realized relative returns and policy outcomes before judging whether China has in fact been a persistently inferior, “largely uninvestable” destination for long‑term growth equity capital over the stated period.

Because the specified timeframe overwhelmingly lies in the future, it is too early to determine whether this prediction is right or wrong.

Jason’s prediction was that, in the months after June 2023, the U.S. government would effectively stop U.S. venture capital from making new investments in Chinese companies, especially in sensitive sectors, and that it would “stop all US venture investing in China.”

What actually happened:

1. Outbound-investment controls were introduced, but only narrowly and not immediately effective.

   • On August 9, 2023, President Biden issued Executive Order 14105 creating an outbound investment program targeting certain Chinese technologies (semiconductors and microelectronics, quantum information technologies, and some AI systems). The order authorizes prohibitions on some investments and notification requirements for others, rather than a blanket ban on all venture investing. (theguardian.com)
   • Treasury’s final regulations for this program were issued October 28, 2024, with an effective date of January 2, 2025—over a year after the prediction window of “coming months” following June 2023. (home.treasury.gov)
   • Congressional and Treasury summaries explicitly describe the program as targeted and narrowly scoped to specific national-security technologies, not as a comprehensive halt to all U.S. investment in China. (congress.gov)

2. U.S. venture investment in China dropped sharply but did not stop.

   • S&P Global data show that U.S.-backed funding rounds in mainland China fell to $3.93 billion across 116 deals in 2023—down more than 50% from 2022, but clearly not zero. (spglobal.com)
   • In the first half of 2024, U.S. private equity and venture capital still invested about $650 million across 45 deals in China, with technology, media and telecom companies remaining the top destination for U.S. private capital in Q2 2024. (spglobal.com)
   • Reporting on the new outbound rules notes that many U.S.-linked investors are restructuring, slowing, or freezing some new China tech investments and racing to comply with the targeted bans on military-related AI/quantum/semiconductor deals—but not that all venture investing has ceased. (ft.com)

3. Net effect vs. the prediction.

   • The government did impose significant outbound-investment controls focused on certain sensitive technologies in China, and these have had a chilling effect on U.S. venture activity.
   • However, they are sector-specific, threshold-based, and partly notification-only, and they came into full legal effect in January 2025, not within just “the coming months” after June 2023.
   • U.S. venture capital activity in China has declined sharply but demonstrably has not been fully or effectively stopped, even in tech sectors broadly defined.

Because the prediction called for the U.S. government to stop all (or effectively all) U.S. venture investing in China within the months after June 2023, and the reality is a delayed, narrow, and partial restriction regime under which investment—though much reduced—continues, the prediction is best classified as wrong.

Available data in 2023–2025 do show a clear trend toward more player‑centric fandom among younger sports fans, including NBA fans. Morning Consult and league executives report that Gen Z is more likely than older cohorts to choose favorite teams based on an individual player, with family/friends still the top factor and geography less important. (pro.morningconsult.com) Broader Gen Z research finds that about 75% identify as fans of specific athletes, 80% follow pro athletes online, and many continue supporting those athletes after they change teams, indicating looser, more player‑driven loyalty patterns. (hardrock.bet) Within basketball specifically, surveys cited in coverage of the NBA’s Gen‑Z strategy note that roughly 40% of Gen Z have a favorite NBA player (more than in other sports) and that LeBron James ranks as their top athlete, reinforcing that individuals, not franchises, anchor much of their emotional attachment. (essentiallysports.com) Industry commentary on next‑gen sports streaming similarly describes younger viewers as increasingly inclined to follow players and personalities rather than teams, consuming more highlights and social content centered on individual stars. (streamingmedia.com)

However, Jason’s prediction is stronger: he says that over time, younger NBA fans will predominantly follow star players rather than teams, and that future viewership and subscription demand will be driven more by specific players than by franchises. On the subscription and rights side, the NBA’s main products and deals in 2024–2025 (League Pass, Team Pass, and national packages like NBA on Prime) are still structured around teams and games, not around “follow this player” subscription SKUs, even though they add more star‑focused shoulder content and personalization. (support.watch.nba.com) There is evidence of a shift in that direction, but we are only about two years removed from the 2023 prediction, and no definitive data yet show that fandom or revenue is predominantly player‑driven rather than team‑driven. Because the claim is explicitly about an eventual, long‑term generational realignment and the current evidence is suggestive but not decisive, the status of the prediction as of November 2025 is best judged as inconclusive (too early to call).

There is currently no real-world implementation of the exact scenario Chamath described—i.e., a market where:

1. Fans can buy standalone, direct-to-consumer subscriptions specifically for one NBA player such as Steph Curry, and
2. Those offerings coexist with separate, purely team-branded DTC subscriptions (e.g., just the New York Knicks) with publicly reported, clearly separated subscriber counts.

What we do have as of 2025:

• The NBA sells League Pass and some Team Pass–style products for access to games, but these are not player-specific subscriptions and detailed per-team subscriber numbers are not public.
• Regional sports networks like MSG+ launched DTC products that include the Knicks, Rangers, etc., but again they are multi-team, regional offerings, not a "Knicks-only" globally scaled subscription with transparent paid-subscriber data.
• There is no evidence of a Steph Curry–only paid streaming subscription (or anything equivalent) with publicly reported multi-million global subscriber figures.

Because Chamath’s prediction is fundamentally about how many subscribers each type of offering (star player vs. team) would attract in a hypothetical market structure—and that structure does not yet exist in a measurable form—the prediction cannot be empirically checked. There are no reliable, disaggregated subscriber metrics for a Knicks-only DTC product, and no Steph-only DTC product to compare against.

Given the lack of a directly testable scenario and absence of relevant public data, the accuracy of the claim ("mid millions for Steph" vs. "a few hundred thousand for the Knicks") cannot be determined, even though enough time has passed.

Therefore the prediction’s status is best classified as ambiguous.

As of late 2025, the NBA has not adopted the specific model Jason described (a single, direct subscription to the NBA for roughly $200/year that gives every game with no ads and bypasses the bundle).

Key facts:

1. Long‑term commitment to traditional media partners: In July 2024 the NBA signed new 11‑year U.S. media agreements with Disney (ABC/ESPN), NBCUniversal (NBC/Peacock), and Amazon Prime Video running through the 2035‑36 season. Games will be spread across broadcast TV, cable, and those companies’ streaming platforms, with the NBA App acting mainly as a “universal access point” that routes fans to those partners—not a single all‑inclusive NBA‑only service. (nba.com) This locks in a rights structure that continues to rely on major broadcasters/streamers rather than shifting to a purely league‑direct model for at least a decade.

2. League Pass remains limited and not truly “all games”: NBA League Pass in 2025‑26 costs about $109.99 per season with ads or $159.99 per season ad‑free in the U.S., and it remains primarily an out‑of‑market product. Nationally televised games and many in‑market games are blacked out and still require access via ESPN/ABC, NBC/Peacock, Prime Video, regional sports channels, or vMVPDs. (yahoo.com) This falls short of the prediction that you could “directly subscribe to the NBA” and “get every game … with no ads” from a single league‑run subscription.

3. Fans still need multiple services, not one league subscription: Current viewing guides for the 2025‑26 season emphasize that U.S. fans need a combination of services—ESPN/ABC, NBC/Peacock, Prime Video, regional outlets plus (optionally) League Pass—to see all games. There is no single, direct‑from‑NBA $200/year product that provides comprehensive, ad‑free access to the entire season. (tomsguide.com)

4. Model is not “bypassing the traditional bundle” in the way described: While there is more streaming and some cord‑cutting friendliness, the economic core is still large rights deals with media companies; many games sit inside their broader bundles or standalone streaming subscriptions. The NBA has not pivoted to making its own DTC subscription the primary, all‑inclusive way to watch the league, and the newly signed contracts strongly suggest that will not be the case through 2035‑36. (nba.com)

Because there is no single, league‑direct, ~$200/year, ad‑free product that gives access to every NBA game and replaces the bundle, the concrete scenario Jason laid out has not materialized. The league has incrementally expanded streaming and DTC options, but the prediction in its stated form is best classified as wrong.

Evidence shows two different outcomes for the two parts of the normalized prediction:

1. Views surpassing 17 million on X/Twitter – This happened quickly. Reports note Tucker Carlson’s first “Tucker on Twitter” episode amassed tens of millions of views, with tweet impressions in the 80–110 million range within about a day, far beyond the initial 17M cited in the podcast discussion.(observer.com) Third‑party analytics (Tubular Labs) estimated roughly 26–27 million actual video views (2+ seconds watched) for episode 1, still well over 17M.(aol.com) So the narrow claim that the initial episode would keep accumulating views beyond 17M was correct.

2. “Larger ongoing distribution on Twitter than he had on Fox News” – This is where the prediction fails.

   • At Fox, Tucker Carlson Tonight averaged about 3.3 million viewers per night in 2022 and roughly 3.25–3.39 million in early 2023, making it one of the highest‑rated shows in cable news.(press.foxnews.com)
   • On Twitter/X, after an initial spike, Tucker on Twitter’s audience fell sharply. Analyses based on Twitter’s own video‑view metric (2+ seconds watched) show episode 1 around 26–27M video views, but by episode 8 views had dropped to about 3.8M, an ~86% decline.(forbes.com)
   • Major media‑measurement experts emphasize that X “views” (brief impressions or 2‑second plays) are not comparable to Nielsen’s average‑minute TV audience; comparisons that claim Carlson’s X audience “dwarfs” his Fox ratings are described as misleading and “apples and oranges.”(aol.com)
   • Longer‑term, coverage of Carlson’s post‑Fox career notes that his overall media reach has diminished relative to his Fox tenure; his subscription Tucker Carlson Network is reported to have on the order of hundreds of thousands of subscribers, far fewer than the multi‑million nightly audience he had on Fox.(theguardian.com)

Because (a) the metrics for Twitter/X and TV aren’t directly comparable, and (b) where we can compare, the best available evidence suggests his sustained, reliably engaged audience is smaller now than it was on Fox, the strong version of the prediction — that he would achieve larger ongoing distribution on Twitter than on Fox News — did not come true.

Overall, weighing both parts together, the more consequential and falsifiable portion (ongoing distribution exceeding Fox) is contradicted by the data and expert assessments, so the prediction is best classified as wrong.

Evidence from multiple analytics sources shows that Tucker Carlson’s Twitter/X show did reach audiences on the order of ~25 million per episode for major releases, matching Jason’s prediction.

Key points:

• Carlson launched “Tucker on Twitter” on June 6, 2023. The first episode’s tweet amassed around 85–120 million tweet views (impressions) within days, indicating extremely wide distribution, even if many of those are just timeline impressions rather than full video views. (newsweek.com)
• More conservative video-view metrics (which require at least ~2 seconds watched) from third‑party analytics (Tubular Labs) and reporters put the first episode around 26–27 million video views, with subsequent early episodes still in the tens of millions before declining. (aol.com)
• Even critics emphasizing that his show later hemorrhaged viewers still cite these initial mid‑tens‑of‑millions video views for the launch episodes, confirming that major releases did, in fact, hit that scale. (independent.co.uk)

Given that Jason’s normalized prediction was that Carlson’s Twitter‑based show would reach an audience on the order of 25 million people per episode or per major release, and early flagship episodes achieved ~25M+ video views and far higher tweet impressions, the prediction is best judged as right (even though that level was not sustained over time).

Only about 2 years have elapsed since the prediction was made in June 2023, while the prediction explicitly concerns a 30–40 year horizon for large-scale migration of “most” U.S. job functions to lower-cost locations using AI tooling and a subsequent reinvention of the U.S. workforce. That kind of structural transformation cannot be reliably evaluated this early. Current data (2023–2025) show rapid AI adoption and some offshoring and reconfiguration of jobs, but not a clear, measurable majority migration of U.S. job functions abroad, nor a completed reinvention of the workforce—nor would we expect to see that yet. Since the time frame for the prediction has barely started, its correctness cannot be judged at this point, so the outcome is inconclusive (too early).

Jason’s claim was that widespread 2023 corporate hiring freezes would be indefinite because it would be easier and cheaper to automate many white‑collar roles with AI than to open and fill new requisitions.

By late 2025, the evidence does not support this:

• U.S. job openings and hires declined but did not freeze. BLS data show openings fell from 10.4 million in January 2023 to 8.9 million in December 2023, with hires and quits continuing throughout the year—i.e., companies kept hiring, just at a cooler pace. (bls.gov)
• Into 2024, job openings declined further to 7.6 million in December, signaling a slowing but still active labor market, not an indefinite freeze. (apnews.com)
• White‑collar hiring specifically is weak enough that some commentators call it a "white‑collar recession," but that still involves ongoing hiring and longer searches, not a standstill. Employers are taking more time and being more selective, not simply automating away all roles. (forbes.com)

On the AI explanation, current research and commentary show AI is one factor among several, not the dominant, immediate reason for corporate hiring restraint:

• Economists and labor‑market analysts emphasize high interest rates and general macroeconomic conditions as primary drivers of reduced white‑collar openings; they explicitly caution that the slowdown so far is “more of an economic story and less of an AI disruption story.” (cnbc.com)
• Studies and forecasts do find substantial future automation potential—Forrester projects generative AI replacing millions of mostly white‑collar jobs by 2030, and Gartner/IPPR estimate a large share of white‑collar tasks could be transformed or replaced. (impactlab.com) But these same analyses note that near‑term, large‑scale job losses are limited by legal, organizational, and reliability constraints.
• Empirical work from Stanford and others shows AI is already reshaping work and disproportionately hitting entry‑level white‑collar roles, but not eliminating broad categories of jobs or ending hiring outright; instead, tasks are being restructured and professionals take on more “AI management” work. (cnbc.com)

Netting this out:

• There is clear ongoing hiring across the economy and within white‑collar sectors in 2023–2025, so hiring freezes were not “indefinite.”
• AI is beginning to substitute for some white‑collar labor and dampen certain junior hiring pipelines, but macroeconomic factors remain the primary, documented cause of the broader hiring slowdown so far.

Because both the permanence (“indefinite freeze”) and the dominant‑cause (“because automating with AI is easier than opening reqs”) parts of the prediction have not materialized as stated, the prediction is best classified as wrong.

As of November 30, 2025, only about 2.5 years have passed since mid‑2023, and Jason’s prediction explicitly allowed up to three years (i.e., roughly until mid‑2026), so the full time window has not elapsed.

There is clear evidence that some ~20‑person AI‑native startups have achieved extremely high output and revenue without large headcount—e.g., Cursor/Anysphere reportedly reached around $100M ARR with roughly 20 employees, and similar small‑team, high‑revenue patterns are noted for other AI startups. (forbes.com) Other examples like Vocal Image (about $12M ARR with a 20‑person team) and ChipStack (a 20‑person AI verification startup acquired by Cadence after demonstrating large efficiency gains) also illustrate very high revenue or productivity per employee. (blog.connfig.com) These cases show that it is possible for 20‑person companies to massively scale output without scaling headcount.

However, system‑level data that would tell us whether “many” 20‑person tech companies have roughly doubled their business output or revenue because of AI does not exist in a clean, aggregated form. Academic and industry studies on AI coding assistants like GitHub Copilot generally find substantial but incremental productivity gains (e.g., on the order of single‑ to low‑double‑digit percentage improvements at the project or developer level, or 30–40% time savings on certain coding tasks), not a universal 2× output figure. (arxiv.org)

Adoption data shows that AI is spreading quickly—e.g., a U.S. Chamber of Commerce report cited in business press finds that a majority of small businesses now use generative AI, and broader surveys suggest a large majority of companies worldwide are using or exploring AI. (kiplinger.com) But these statistics don’t specify 20‑person tech companies, nor do they quantify how many have doubled output at fixed headcount.

Because (1) the prediction’s 3‑year horizon has not yet fully expired, and (2) available evidence shows notable supporting anecdotes but no comprehensive data to confirm or refute that many such companies have doubled output, the prediction cannot be definitively graded yet. It remains plausible but unproven, so the status is best described as inconclusive (too early).

As of late 2025, there are clear signs that AI is tightening white‑collar and entry‑level hiring, but the specific scenario Jason described – large swaths of workers being unable to get interviews for anything except in‑person service jobs – is not yet clearly established.

Evidence that points toward his concern:

• Several analyses find a steep drop in entry‑level and junior white‑collar opportunities since the spread of generative AI. Revelio Labs estimates U.S. entry‑level postings are down ~35% since January 2023, with AI identified as a major factor.(cnbc.com) The “jobpocalypse” coverage similarly reports entry‑level offers in the U.S. and U.K. falling by about one‑third and notes that many employers say AI lets them cut staff.(en.wikipedia.org)
• A summary of American Staffing Association data states that roughly 40% of white‑collar job seekers in 2024 failed to secure any interview, indicating serious bottlenecks for many aspiring knowledge workers.(medium.com)
• Multiple studies and forecasts (IMF, WEF, etc.) highlight that AI’s largest exposure is in high‑skill, white‑collar occupations and that many employers expect to reduce headcount where AI can automate tasks.(institute.global) This supports Jason’s directionally pessimistic view about knowledge‑worker competition.

Evidence against saying his prediction has already come true:

• Overall labor‑market data do not show masses of workers being employable only in service roles. U.S. unemployment in 2025 is elevated but not catastrophic (around the mid‑4% range), with weakness and selectivity in white‑collar hiring but ongoing hiring across many professional sectors.(apnews.com) College‑graduate unemployment has risen (to about 5.8% in 2025) but remains far below levels that would indicate wholesale exclusion from non‑service work.(en.wikipedia.org)
• Studies of AI adoption emphasize that its labor effects are still in early stages. One 2025 synthesis notes that as of 2024 only about 5% of U.S. businesses had implemented AI solutions and that 63% of workers reported little or no AI use at work, suggesting that broad, AI‑driven restructuring is still nascent.(dataconomy.com) Another large analysis of job postings and skills exposure finds extensive task reshuffling but very few skills that are fully automatable today, arguing against immediate mass displacement.(siai.org)
• There is evidence of some young people proactively shifting into trades and manual/blue‑collar work (construction, HVAC, etc.) partly because white‑collar entry paths look worse in an AI era, but this is not the same as being able to get interviews only for service jobs; it reflects both choice and a tougher white‑collar market.(businessinsider.com)

Given that the podcast aired in mid‑2023 and Jason spoke about what would happen “over the coming years,” we are only about two and a half years into that window. AI is clearly contributing to a squeeze on junior and some knowledge‑worker roles, and some data (e.g., high shares of white‑collar job seekers getting no interviews) are directionally consistent with his worry. But we do not yet see solid evidence that a large segment of the workforce can only get interviews for in‑person service work, nor that AI adoption has matured enough to settle the long‑run outcome. For now, the prediction’s truth value remains inconclusive rather than clearly right or wrong.

As of November 30, 2025, there has been no realized catastrophic or existential AI event (e.g., human extinction or near‑extinction, loss of global human control, or comparable catastrophe). Surveys and discussions of AI existential risk focus on future probabilities (e.g., risk by 2100 or over the next several decades), not on events that have already occurred.(en.wikipedia.org)

On the employment side, the observed economic impact has been disruptive but gradual, not an overnight collapse:

• The New York Fed reports that, so far, AI adoption "has not yet led to major job losses" and that near‑term labor‑market effects have been modest.(reuters.com)
• OECD analysis of AI and jobs finds that to date AI has had little effect on aggregate employment, with adoption still relatively low and many firms adjusting slowly (via attrition, retraining, or partial task automation).(oecd.org)
• A 2025 Goldman Sachs analysis likewise concludes that generative AI has not yet produced discernible changes in unemployment or productivity at the macro level, implying that the feared rapid jobs shock has not materialized.(fortune.com)
• Other recent reviews similarly emphasize that an "AI jobs apocalypse" has not arrived; displacement is emerging but is slow, uneven, and concentrated in specific sectors, consistent with a drawn‑out transition rather than a sudden crash.(siai.org)

In short, between June 2023 and late 2025, no sudden AI doomer scenario has actually unfolded, and the disruptive impacts we do see are incremental and evolving over years, matching Sacks’s claim that such scenarios would not "play out overnight" and that the real effects would "take a while." Therefore the prediction is right given the evidence up to the evaluation date.

Evidence up to November 2025 shows that AI is significantly changing radiology workflows but has not made radiology a poor or shrinking career, while the broader physician workforce remains in shortage.

Key points:

• Radiology is being augmented, not gutted, by AI. Reporting from 2025 notes that over two‑thirds of U.S. radiology departments use AI tools and that more than 75% of FDA‑cleared medical AI products are for radiology, but experts emphasize these systems mainly triage cases, flag findings, or help with reporting rather than replace radiologists. (washingtonpost.com)

• Radiologist demand remains high and is projected to stay that way for decades. Multiple workforce studies (Neiman Health Policy Institute and related analyses) project that today’s radiologist shortage will persist through 2055 as imaging demand grows roughly in line with or faster than radiologist supply, meaning the shortage neither disappears nor worsens dramatically without intervention. (drugs.com) Job‑market discussions similarly describe a nationwide radiologist shortage, long hiring times, and many open positions—conditions inconsistent with radiology being a “bad” or oversupplied specialty. (medscape.com)

• Doctors overall are clearly still in demand. Updated projections from the Association of American Medical Colleges and related summaries in 2024–2025 forecast a U.S. physician shortfall of up to ~86,000 doctors by 2036 (and higher in some federal estimates), spanning primary care and many specialties. These reports explicitly frame physician shortages—not AI displacement—as the dominant workforce problem. (ama-assn.org)

Taken together, AI has indeed begun to automate and transform parts of radiology, but it has not reduced the specialty to a poor career choice; on the contrary, radiologists are scarce and heavily recruited. Meanwhile, medicine in general remains viable and in high demand. Because the central claim that AI would make radiology a bad career has been contradicted by current data, the overall prediction is best classified as wrong.

Available evidence as of November 30, 2025 does not show that long‑haul truck driving jobs have been significantly reduced by self‑driving technology yet, although there is clear activity and investment that could threaten those jobs in the future.

What we see today (late 2025):

1. Autonomous trucking is in limited pilots, not at mass displacement scale. Companies like Aurora, Kodiak, Torc (Daimler), and others are running autonomous truck pilots on select highway lanes in the U.S., generally with safety drivers or close human supervision, and are working toward commercial launches around the mid‑2020s, but these are not yet replacing a large share of long‑haul drivers nationwide. (This is widely reported across transportation and business press; there is no credible data showing large‑scale job loss specifically attributable to autonomous trucks.)

2. Truck driving employment remains high. Public labor statistics and industry commentary through 2024–2025 indicate that trucking companies are still hiring and frequently report driver shortages or tight labor markets, not a collapse in demand for long‑haul drivers due to automation. There is no major report of self‑driving trucks causing substantial net job losses among long‑haul drivers at the national level.

3. Transportation and logistics companies obviously continue to exist. Large carriers (e.g., Knight‑Swift, J.B. Hunt, Schneider, FedEx, UPS, etc.) and countless regional and local fleets are still operating. That portion of the prediction is trivially correct but wasn’t really in doubt.

Why the verdict is ‘inconclusive’:

• The prediction is about future structural impact ("I wouldn’t want to be a truck driver… because of self‑driving"), not explicitly tied to a short, dated horizon like 2024 or 2025. It implies a long‑term threat over a multi‑year/decade timescale.
• As of late 2025, self‑driving trucks are credible and progressing but have not yet produced the “significant reduction” in long‑haul truck driving jobs the quote anticipates. At the same time, the technology is clearly advanced enough that it could plausibly threaten those jobs later.

Because (a) no large‑scale displacement has occurred yet, but (b) the stated prediction is long‑term and does not specify a deadline, the fairest assessment as of November 30, 2025 is that it is too early to say whether it will ultimately be right or wrong.

As of late 2025, both the legal and accounting professions remain large, mainstream job categories, with evidence that AI is changing task mix rather than eliminating the roles.

In law, U.S. legal-sector employment has been rising, with job numbers in 2025 approaching prior peaks even as firms increase their use of AI to streamline work. (reuters.com) Surveys of legal professionals and AI adoption show a growing consensus that generative AI is a powerful tool but not a replacement for lawyers; many legal professionals explicitly state that AI will augment their work, not eliminate lawyer roles. (globallegalpost.com) Consistently, the U.S. Bureau of Labor Statistics projects lawyer employment to grow about 4% from 2024–2034, indicating continued demand for human attorneys even as some routine work is automated. (bls.gov)

In accounting, AI and automation are being heavily adopted but framed as productivity tools. Industry reports describe AI automating repetitive tasks (data entry, basic analysis) while highlighting that human judgment, ethics, and client-facing advisory work remain central; experts and professional bodies stress that GenAI is meant to support, not replace, accountants. (at-mia.my) Surveys in 2024–2025 find most tax and accounting firms expect significant GenAI integration but only a small minority now view it as a threat to employment, and many firms still report talent shortages rather than surplus. (cfobrew.com) BLS projections show employment of accountants and auditors growing around 5–6% over the coming decade, with the agency explicitly noting that while some tasks will be automated by AI and related tech, overall demand for accountants is expected to rise and their work to shift toward higher-value analysis and advisory duties. (bls.gov)

Some narrower, more routine roles (e.g., bookkeeping and clerical accounting positions) are projected to decline as automation spreads, but that is materially different from eliminating the broad professional categories of lawyers and accountants, which remain in demand and are forecast to grow. (bls.gov) Given that, about 2.5 years after the prediction, the best available data and expert views align with Sacks’s claim that these professions are sufficiently general and judgment-intensive that AI is unlikely to eliminate them; instead, they are evolving into AI-augmented roles. This supports classifying the prediction as right (so far).

As of late 2025, available evidence broadly supports Sacks’s prediction that (1) AI tools are making lawyers and accountants more productive and (2) this has not translated into a clear reduction in overall demand for these professionals.

1. Productivity gains are real and measurable

• A 2024 Thomson Reuters Future of Professionals report surveying legal, tax, and accounting professionals finds they expect AI to save about 4 hours per week in the near term and up to 12 hours per week within five years, with 77% saying AI will have a high or transformational impact on their work. (thomsonreuters.com)
• Multiple 2025 surveys of legal professionals (e.g., Everlaw’s Ediscovery Innovation Report) report that nearly half of respondents are already saving 1–5 hours per week with generative AI, with heavy users reclaiming up to ~260 hours per year (about 32.5 working days). (everlaw.com)
• A 2024–25 St. Louis Fed analysis of U.S. workers using generative AI finds average time savings of 5.4% of work hours among users, confirming that genAI yields nontrivial productivity gains at the worker level. (stlouisfed.org)

2. Lawyer employment and demand have not fallen

• U.S. Bureau of Labor Statistics (BLS) data show that lawyers held about 859,000 jobs in 2023 and about 864,800 jobs in 2024—an increase, not a decline, during the initial phase of wide genAI adoption. (bls.gov)
• A 2025 BLS Monthly Labor Review article that explicitly incorporates generative AI effects projects that overall legal occupations will grow 3.7% from 2023–2033, with lawyers specifically projected to grow 5.2%, even while acknowledging that genAI can substantially increase productivity in legal research and document review. (bls.gov)
• Industry data show that demand for legal services is rising: a 2025 LawCrossing analysis reports that legal demand at firms grew 2.8% in 2024, the fastest growth since 2021, with broad-based increases across litigation and corporate work. (lawcrossing.com)
  These patterns match the prediction: AI is boosting productivity, yet law remains a growth field rather than one clearly shedding lawyers because of AI.

3. Accounting: productivity plus shortage, not surplus

• BLS data show accountants and auditors held about 1,538,400 jobs in 2022 and about 1.6 million jobs in 2024, with projections of roughly 5% growth from 2024–2034 (1,579,800 to 1,652,600 jobs). (scribd.com)
• The BLS Occupational Outlook Handbook explicitly states that while platforms such as cloud computing, AI, and blockchain will automate routine accounting tasks and increase efficiency, this change is not expected to reduce overall demand for accountants; instead, their advisory and analytical duties are expected to become more prominent. (bls.gov)
• A 2025 CPA Journal article, drawing on BLS projections, notes that accounting and auditing jobs are expected to grow about 5.8–6% from 2023–2033 and argues that this job growth is tied to new roles that require collaboration between humans and AI models. (cpajournal.com)
• At the same time, several analyses describe a shortage of accountants: CPA Journal and Fortune both cite BLS-based estimates that the U.S. accounting workforce has shrunk by roughly 300,000–340,000 over the past five years, mainly due to retirements and fewer new entrants, not because AI eliminated the need for accountants. CFOs report difficulty hiring and are investing in AI and automation partly to cope with this shortage, while still emphasizing the need for experienced human talent. (cpajournal.com)
  This is important context: the total headcount of accountants has been pressured downward by demographic and pipeline issues, but the underlying demand remains strong or growing; AI is being used to stretch scarce talent, not because firms no longer “need” accountants.

4. Localized AI-related cuts exist but don’t overturn the aggregate picture

• Some high-profile firms have cited AI when restructuring: for example, Clifford Chance is cutting about 10% of its London business-services workforce (finance, HR, IT), partly due to increased AI use, and PwC has abandoned a global plan to add 100,000 staff by 2026, cutting 5,600 jobs and reducing graduate hiring while simultaneously investing heavily in AI and seeking large numbers of technologists. (theguardian.com)
• U.K. employer surveys from the CIPD show that a minority of large firms in sectors like legal, accounting, and consultancy anticipate AI-driven staff reductions, especially in junior and support roles. (ft.com)
  These examples show that AI can reduce staffing in specific firms or support functions, but they do not amount to a broad, AI-driven collapse in employment for lawyers and accountants as occupations.

Overall assessment

• The core of Sacks’s prediction was that as AI tools improve, lawyers and accountants would “be able to get more done” but that this would not necessarily mean “we’ll need less of them.”
• Current evidence shows (a) clear productivity gains from AI for legal and accounting professionals, and (b) continued or projected growth in overall employment and demand for lawyers and accountants, with shortages and pipeline issues in accounting driven mainly by demographics and career-choice trends rather than by AI substitution. BLS explicitly expects AI-enabled efficiency to change the nature of accounting work, not to reduce overall demand, and projects positive growth for both accountants and lawyers under scenarios that factor in AI. (bls.gov)
  Given the time elapsed (roughly 2.5 years since mid-2023) and the available data, Sacks’s view matches the direction of observed trends: AI has boosted productivity without clearly reducing the total “need” for lawyers and accountants at the occupational level. It remains possible that long-run dynamics will evolve, but as of November 30, 2025, the prediction aligns with the evidence well enough to count as right (so far) rather than inconclusive.

As of 30 November 2025, Article II, Section 1, Clause 5 of the U.S. Constitution still requires the president to be a “natural born Citizen,” and no constitutional amendment has been ratified to change that requirement or to allow non–natural-born citizens to become president.(constitution.congress.gov)

Chamath’s prediction has two parts:

1. Short-term claim: “It can't happen in the next few years.” So far (about 2.5 years after June 2023), this has held true—no such amendment has passed—but we have not yet reached the end of a reasonable “few years” window.
2. Long-term claim: A small probability that such a change could occur in roughly 25–30 years from 2023. That window runs out around 2048–2053 and is entirely in the future.

Because the core of the prediction concerns events up to several decades out, and we are still near the beginning of that horizon, it is too early to determine whether the prediction is ultimately right or wrong.