Trump: Send National Guard to SF, China Rare Earths Trade War, AI's PR Crisis

An in‑person Trump–Xi meeting in South Korea did happen on October 30, 2025, in Busan, on the margins of the APEC South Korea summit. At that summit they agreed to cut the U.S. fentanyl‑linked tariff on Chinese imports from 20% to 10% (bringing overall duties on most Chinese goods down from about 57% to 47%) and for China to resume large U.S. soybean purchases and allow rare‑earth exports again. (apnews.com)

However, the deal was explicitly framed as a limited, one‑year truce rather than a comprehensive settlement. HSBC described it as a "one-year tactical trade truce" that temporarily reduces one specific tariff, pauses some reciprocal tariffs and shipping measures for a year, and suspends China’s new rare‑earth export controls for a year, while leaving fundamental disputes over tariffs, overcapacity, and tech export controls unresolved. (privatebanking.hsbc.com) Foreign Policy likewise called it a "framework trade agreement" that institutes a one‑year pause on further hostilities but leaves earlier, very high tariffs and other export controls largely in place. (foreignpolicy.com)

Analysts and reporters were clear that this was not a sweeping, structural reset of U.S.–China trade. The Washington Post characterized it as a trade truce that "avoids an open rupture" but mainly de‑escalates without solving core issues, noting that most tariffs remain and that the relationship is still drifting apart. (washingtonpost.com) The Foreign Exchanges newsletter was even more direct, writing that it "would be wrong to call this a comprehensive trade deal" because it barely touches the existing tariff wall and leaves structural problems untouched. (foreignexchanges.news)

Because the Busan summit produced a narrow, time‑limited truce and framework rather than the kind of comprehensive, "grand" trade deal the prediction envisioned, the prediction is best judged as wrong.

Chamath’s prediction on Oct 17, 2025 was that local political and community opposition that blocks or forces withdrawal of large AI/data-center projects would continue and evolve into a broader trend “in the coming years,” rather than remain isolated incidents.

What has happened so far (before and after the prediction):

• Before the podcast (trend already emerging):

  • An NPR/10a Labs analysis found that 16 data‑center projects nationwide were delayed or rejected between May of the previous year and March 2025 due to community resistance, framing it as mounting nationwide opposition to AI/data centers, not just one‑off cases. (wboi.org)
  • Google withdrew a 468‑acre data center project in Indianapolis after intense local pushback—hundreds of residents packed hearings and celebrated when Google pulled its rezoning request. (axios.com)
  • In Tucson, the City Council unanimously rejected the massive Amazon‑linked Project Blue data center after weeks of protests and organizing by groups like No Desert Data Center, explicitly citing water, energy, and transparency concerns. (azluminaria.org)

• After the podcast (Oct 17–Nov 30, 2025):

  • Microsoft canceled a 244‑acre data center (Project Nova) in Caledonia, Wisconsin following pushback from residents and local officials, stating it would seek a site better aligned with community needs. (tomshardware.com)
  • A proposed $17 billion “Project Sail” data center in Coweta County, Georgia has triggered large‑scale opposition, with a 3,600‑member citizen group (“Stop Project Sail”), public meetings, and widespread concern over water use, noise, and construction impacts. (ft.com)
  • In Springdale Borough, Pennsylvania, residents and environmental groups are opposing a 565,000‑square‑foot AI data center at the site of the former Cheswick Generating Station, citing pollution, noise, and high electricity and water use; the borough council’s final decision is still pending. (axios.com)

These examples—spanning the Midwest, Southwest, Southeast, and Northeast—show that new cases of community backlash and even project cancellations have indeed occurred since the prediction, and they fit the pattern Chamath described.

Why the verdict is still “inconclusive (too early)” rather than “right”:

• The prediction explicitly concerns a multi‑year trajectory (“the beginning of a trend” that will continue “in the coming years”). As of today (Nov 30, 2025), only about six weeks have passed since the podcast.
• While post‑prediction events in Wisconsin, Georgia, and Pennsylvania are consistent with his claim and reinforce that this is more than a few isolated incidents, we cannot yet know whether this level of opposition will persist or intensify over “the coming years.”

So: early evidence strongly aligns with Chamath’s thesis that local opposition to AI/data centers is spreading and affecting projects across multiple states, but because his forecast was explicitly about a long‑term trend over years, not just the next few months, there hasn’t been enough time to say definitively that the prediction has come true. Hence the result is inconclusive (too early) rather than clearly right or wrong.

The prediction is explicitly open‑ended – “all those jobs are going away. It’s just a matter of when” – with no concrete time horizon. As of November 30, 2025, there is strong evidence that these roles still exist at large scale in the U.S., but that does not by itself falsify a claim that they will disappear eventually.

Evidence that the jobs still exist in large numbers (so they are not “gone” yet):

• Uber / DoorDash‑type driving: Uber reports 8.8 million drivers and couriers on its platform worldwide in Q2 2025, with U.S. rideshare drivers estimated at over 1 million, and total U.S. rideshare drivers (Uber + Lyft) around 1.7 million as of 2025. These numbers have been growing, not shrinking to zero. (grabon.com)
• Truck drivers: The U.S. Bureau of Labor Statistics (BLS) reports about 2.24 million heavy and tractor‑trailer truck driving jobs in 2024 and projects 4% employment growth from 2024–2034, roughly in line with average occupation growth. (bls.gov) This is inconsistent with near‑term elimination.
• Amazon warehouse/factory workers: Amazon’s total global workforce was about 1.5–1.6 million employees in 2024–2025, most in warehouse and logistics roles, and the company is investing heavily in robotics and AI while also continuing large‑scale human operations. (ft.com) Internal documents suggest a goal to automate ~600,000 U.S. warehouse jobs and 75% of operations by 2033, which implies gradual reduction rather than completed elimination today. (nypost.com)

Why the status of the prediction is inconclusive rather than wrong:

• The quote gives no deadline (e.g., “within 5 years”). It just asserts that these jobs will go away “over time” and that it’s “a matter of when.” That kind of unbounded forecast can’t be disproven only ~6 weeks after it was made.
• Available data show substantial current employment and even projected medium‑term growth for truck drivers, plus ongoing large workforces for rideshare and Amazon warehouses. That undermines any short‑term version of the claim, but is not enough to prove that long‑run elimination by automation and AI won’t happen.
• Because the horizon is undefined and we cannot yet observe the long‑run outcome, the correct assessment as of November 30, 2025 is that it’s too early to judge the prediction’s ultimate truth.

So, the prediction is not yet supported by current facts, but it also hasn’t been falsified, given its vague, long‑term wording. Hence the status: inconclusive (too early).

The prediction is explicitly long‑term: it concerns what will happen “over the coming years and decades” to employment, productivity, and living standards as AI diffuses. Only about six weeks have passed since the podcast’s release (17 Oct 2025 → 30 Nov 2025), far too short to judge decades‑scale claims.

Current evidence is mixed and still emerging:

• Multiple studies (Yale–Brookings, U.S. Census ABS, Fed and think‑tank reports) find that, so far, AI adoption has not caused large net job losses at the aggregate level; most firms report little change in worker numbers, with some retraining and localized displacement instead of mass unemployment. (theguardian.com)
• At the same time, there are meaningful pockets of disruption: higher unemployment and reduced hiring in AI‑exposed, especially entry‑level, white‑collar roles; job cuts explicitly attributed to AI; and concerns about a “jobpocalypse” for new graduates. (en.wikipedia.org)
• Research broadly agrees that AI can generate significant productivity gains at the task and firm level and may already have added a measurable boost to aggregate labor productivity, but the extent to which this will translate into widely shared higher living standards remains uncertain. (stlouisfed.org)

Because (1) the forecast horizon is many years to decades, (2) we are still in the very early stages of widespread AI deployment, and (3) the long‑run balance between displacement, new job creation, productivity gains, inequality, and living standards is not yet observable, there is not enough elapsed time or evidence to classify the prediction as right or wrong. It remains a plausible but untested long‑term scenario rather than a claim that can already be verified or falsified.

The prediction explicitly concerns “the foreseeable future (at least over the coming decade)” starting from the episode’s release on 17 October 2025, so it cannot reasonably be judged less than two months later (by 30 November 2025). A 10‑year labor‑market claim needs most or all of that horizon to play out.

Early evidence about AI and jobs is still mixed and preliminary:

• Surveys and regional data (e.g., New York Fed) find AI adoption rising but layoffs directly attributed to AI remain rare; many firms report using AI mainly as a productivity tool and plan to retrain workers, which aligns with the “AI as tool/augmentation” part of the claim, at least so far. (pymnts.com)
• Large workforce studies (e.g., SHRM 2025) show substantial task-level automation but relatively few jobs that are both highly automatable and free of non‑technical barriers to full replacement, suggesting transformation and partial automation rather than wholesale job loss for now. (shrm.org)
• Other analyses (Goldman Sachs Research, PwC, WEF and similar summaries) project that AI may eventually displace a meaningful share of jobs, but also emphasize uncertainty and the potential for offsetting job creation and productivity gains over many years, not weeks. (goldmansachs.com)
• Prominent experts like Geoffrey Hinton and Dario Amodei warn that AI could cause very large job losses by 2030, while others remain skeptical that such extreme outcomes are likely, underscoring that long‑term impacts remain unsettled. (news.com.au)

Because (1) the prediction is explicitly decadal, (2) only a tiny fraction of that period has elapsed, and (3) forward‑looking analyses still disagree about whether AI will mainly augment or substantially replace workers, there is not yet enough evidence to label the prediction as right or wrong. It is therefore too early to call.

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

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

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

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

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