Featuring
The original quartet takes on three hard questions: whether AI is already displacing workers, whether one cool inflation print really opened a golden age, and whether China's chip push is copying ASML or building a broader route around it. Jason and Sacks bring the most heat, while Friedberg has the episode's best technical correction: a chokepoint is not a strategy if your rival is redesigning the road.
Spice rack
Did the December 2025 data justify calling inflation solved and a 2026 golden age imminent?
Original point: A 2.7% CPI print, private-sector job growth, lower government payrolls, strong GDP, and coming tax cuts put the country on the cusp of a golden age.
What everyone argued
Chamath Palihapitiya
Chamath says the direction is positive because private jobs are growing, federal employment is shrinking, and fewer regulatory layers should lower the cost of building in America. He treats the data as a foundation for productivity rather than proof that households already feel relief.
Jason Calacanis
Jason argues that lower inflation is not lower prices, unemployment had risen, and many households outside the equity boom still felt squeezed. He says calling it a golden age ignored the gap between campaign promises, official aggregates, and public experience.
David Sacks
Sacks says the November report beat expectations, core inflation was at a multi-year low, federal departures explained much of the payroll weakness, and tax cuts plus AI investment pointed to a gangbusters 2026. He treats household skepticism as a lagging snapshot.
David Friedberg
Friedberg says two parts of the administration's 3-3-3 framework were on track: growth above 3% and inflation below 3%, while deficit reduction remained unfinished. He expects those top-line gains to reach wages, jobs, and affordability but calls that transmission TBD.
Winner circle
Jason wins on calibration and hindsight. Sacks correctly identified real late-2025 strengths, while Chamath and Friedberg offered more careful versions of the bullish case. But 'solved' and 'gangbusters' carried a burden the 2026 inflation and payroll data did not meet; Jason's mixed picture was closer to what followed.
Commentary
Chamath Palihapitiya
Assumptions and fact checks
A smaller federal workforce will materially reduce regulatory cost and raise productivity.
Why it mattersHeadcount cuts can remove delay, but outcomes depend on which roles disappear, whether rules change, and whether work shifts to state agencies or private compliance teams.
Jason Calacanis
Jason has the right distinction—slower price growth is not cheaper groceries—and the better time horizon. His '15%' unemployment line is technically defensible but rhetorically inflated; percentage points would have been cleaner.
Assumptions and fact checks
Poor economic approval and continuing affordability pressure were better guides to the near-term outlook than one favorable inflation release.
Why it mattersSentiment is not a forecast by itself, but the November release had shutdown-related data gaps and later inflation and payroll data vindicated caution about extrapolating from a single print.
Prices were still rising even though the inflation rate had fallen.
CheckThe CPI-U was 2.7% higher than a year earlier in November 2025. Disinflation means prices rise more slowly; it does not mean the overall price level falls.
The unemployment rate had risen 15% from 4.0% to 4.6%.
CheckThe arithmetic is correct as a relative increase, and BLS reported 4.6% in November. The clearer description is a 0.6-percentage-point rise; '15%' sounds larger and obscures that unemployment remained moderate.
David Sacks
Sacks has several good inputs and an unjustified certainty multiplier. The episode's most important analytical error is turning a noisy, shutdown-affected inflation release into 'solved' and then turning 'solved' into a full-year boom.
Assumptions and fact checks
Inflation was essentially solved and would keep falling through 2026.
Why it mattersOne favorable year-over-year print did not establish persistence. Headline CPI later rose to 3.5% in June 2026, while core was 2.6%.
Tax cuts and AI capital spending would produce a gangbusters 2026 labor market.
Why it mattersJuly payrolls fell slightly and prior months were revised sharply lower. Unemployment remained low, but the available 2026 data do not describe a broad hiring boom.
November 2025 CPI was 2.7% and core CPI was 2.6% year over year.
CheckBLS reported both figures. October price data were unavailable because of the appropriations lapse, so the release did not support a normal one-month trend calculation.
The October federal payroll drop was 162,000 as deferred-resignation participants left payrolls.
CheckBLS reported a 162,000 October decline and explicitly connected it to federal employees who accepted deferred resignation offers.
David Friedberg
Friedberg earns credit for the episode's cleanest hedge: the macro dashboard can improve before Main Street does. His framework is useful as a checkpoint, not as proof that the household economy has arrived.
Assumptions and fact checks
Meeting top-line growth and inflation targets would soon flow through to jobs, affordability, and wages.
Why it mattersThe transmission is plausible but neither automatic nor evenly distributed. Friedberg appropriately marked it TBD, and subsequent payroll and inflation data remained mixed.
Real GDP growth was above 3% in the latest reported quarter.
CheckBEA's initial estimate put third-quarter 2025 real GDP growth at a 4.3% annual rate.
Does current evidence show AI is already displacing workers, or only that specific jobs face future risk?
Original point: Warnings about present AI job loss retreat from a dramatic, testable claim to an unfalsifiable future prediction when aggregate labor data show no broad disruption.
What everyone argued
Jason Calacanis
Jason says he has consistently warned about displacement, not economy-wide collapse. He points to robotaxi operators reducing driver recruitment in deployment markets and argues that companies are already planning transitions for workers whose categories will shrink.
David Sacks
Sacks argues that the measurable present-tense claim fails: Vanguard found faster job and real-wage growth in highly AI-exposed occupations, and Yale found no discernible economy-wide disruption. He concedes that some categories will lose jobs, but says the relevant question is whether new jobs more than offset them.
Winner circle
Sacks wins the evidence round, narrowly. He states the testable aggregate question, cites the best available broad studies, and concedes that particular jobs can disappear. Jason is right that distribution matters, but his examples establish exposure and transition planning—not the scale of current net displacement.
Commentary
Jason Calacanis
Jason's strongest point is about distribution, not totals: a flat national number can hide real losses for drivers or junior workers. He weakens it by sliding from private examples to an industry-wide diagnosis without quantifying how many workers are affected.
Assumptions and fact checks
Robotaxi deployment and corporate automation plans are enough to describe present AI displacement as a material labor-market problem.
Why it mattersThey establish a credible category-level risk, but materiality depends on scale, timing, worker transitions, and jobs created elsewhere. Anecdotes from two cities cannot settle the national net effect.
David Sacks
Sacks wins the narrow present-tense question but labels too much a hoax. Aggregate evidence answers whether the whole market has moved; it does not erase early losses inside particular occupations or the measurement limits Yale itself emphasizes.
Assumptions and fact checks
Faster growth in AI-exposed occupations means AI is raising worker productivity and labor demand.
Why it mattersThe mechanism is plausible, but exposure groups have different education, industry, and business-cycle profiles. Vanguard's comparison alone cannot identify causation.
Vanguard found annualized post-2023 job growth of 1.7% and real-wage growth of 3.8% in highly AI-exposed occupations, versus 0.8% and 0.7% elsewhere.
CheckThose values appear in Vanguard's 2026 outlook, based on data through August 2025. They are descriptive comparisons, not a causal estimate of AI's effect.
Broad labor-market research had found no discernible economy-wide disruption attributable to AI.
CheckYale's Budget Lab reported no clear aggregate effect in 2025 and, using a synthetic differences-in-differences design, still found no statistically or economically significant employment or wage effect through early 2026.
Is China's lithography push mainly an ASML reverse-engineering project, or a broader independent path?
Original point: Calling the reported prototype stolen ASML technology is too narrow because China has funded a decade-long program pursuing domestic tools and alternative lithography methods.
What everyone argued
Chamath Palihapitiya
Chamath assumes China has copied EUV but says the strategic damage may be limited because future inference chips can move toward memory-centric architectures built on less advanced nodes, where U.S. software and compilers remain stronger.
David Sacks
Sacks calls ASML's EUV monopoly the West's cleanest chokepoint and says reverse engineering is China's obvious first move. He also notes that China has pushed DUV much further than expected and can offset weaker chips with networking and scale.
David Friedberg
Friedberg argues that China is solving the manufacturing problem, not merely chasing an ASML replica. He cites state-backed funds, Tsinghua research on AI-assisted inverse lithography, and alternative optics and process methods as evidence of a primacy campaign.
Winner circle
Friedberg wins the framing, with a large timing penalty. China's effort plainly extends beyond one reverse-engineered EUV machine, and later domestic DUV production strengthens that point. Sacks is right that copying the proven route is rational, while Chamath is right to focus on delivered systems; none of that supports Friedberg's near-term production confidence.
Commentary
Chamath Palihapitiya
Chamath asks the right commercial question—what workload the system delivers—but leaps from an architectural option to a strategic cushion. The bottleneck can move without disappearing.
Assumptions and fact checks
Memory-centric inference architectures will reduce the strategic value of cutting-edge process nodes.
Why it mattersArchitecture and software can trade compute for memory, but bandwidth, power, packaging, yield, and total cost still matter. No public result yet establishes that older nodes erase the advanced-node advantage.
David Sacks
Sacks is right that copying and innovating are complements: a state program can pursue the proven machine first while funding alternatives. That concession actually moves him closer to Friedberg's broader-system framing.
Assumptions and fact checks
Copying a working ASML-style EUV system is China's fastest path to closing the advanced-node gap.
Why it mattersA proven architecture lowers technical risk, and the reported use of former ASML engineers supports that path. It can coexist with alternative domestic research rather than excluding it.
David Friedberg
Friedberg wins the map and loses the stopwatch. He is right that China is building an ecosystem, not just photocopying one machine; his cited paper, however, is a computational-lithography review, not proof that the hardware gap is closed.
Assumptions and fact checks
China is likely to field Huawei-linked production lithography on Friedberg's 2026-2027 timeline.
Why it mattersDomestic DUV production is progress, but Reuters said the reported EUV prototype had not produced working chips and sources placed that milestone around 2028-2030. Public evidence does not support the faster timeline.
The cited Tsinghua paper documents AI methods that improve computational and inverse lithography.
CheckThe peer-reviewed review covers AI-assisted modeling, source-mask optimization, thick-mask simulation, and pattern fidelity. It does not by itself demonstrate a complete alternative EUV scanner or working chips.
China's program was broader than one EUV prototype and included domestic lithography equipment development.
CheckReuters reported in July 2026 that a state-backed Chinese company had begun producing domestic immersion-DUV machines by combining teams from several local equipment efforts. The machines still required testing and lagged ASML.
Chamath keeps his claim directional, which ages better than declaring victory. The missing link is evidence that staffing cuts changed approval times or investment rather than merely moving the bottleneck.