All-In Podcast

Multiple episodes converge on the same number: the US is on track to be short several "Californias" worth of electricity by 2050, before AI demand is even added, and every hyperscaler is now racing to lock down power any way it can - nuclear restarts, mobile turbines, or orbit. - 1 gigawatt of AI compute needs roughly 50,000 tons of copper and, if powered entirely by solar, more land than the city of San Francisco (2026-06-10) - Microsoft got Constellation to restart the Three Mile Island reactor with a 20-year, above-market $100/MWh price floor because supply, not capital, is now the binding constraint (2026-06-12) - Chamath projects a US energy shortfall equal to 2.5-3 Californias by 2050, evidenced by a PJM capacity auction that sought 7-8 GW and drew only ~150 MW of bids (2026-07-11, 2026-07-18) - Gelsinger argues energy capacity is a natural governor that keeps the AI buildout from becoming an unconstrained bubble, since nobody can build data centers faster than the grid can supply power (2026-07-15) - Space-based data centers become cost-competitive once launch costs fall to $200-300/kg (versus ~$1,000/kg today), because orbital solar collects ~5x more energy than the same panel on the ground with no battery needed (2026-06-06)

Copper, silver, and DRAM are each running structural, multi-year supply deficits that the panel argues the market hasn't fully priced, with China holding a processing (not just extraction) advantage that a US mining boom alone can't close. - Meeting even GDP-trend copper demand over the next 18 years requires mining as much copper as humanity has mined in the last 10,000 years combined, and the rare-earth bottleneck is China's processing know-how, not raw scarcity (2026-06-10) - Silver has roughly a 3-year runway before above-ground inventory runs out, given a 200-million-ounce annual deficit (2026-06-10) - DRAM, not GPUs or power, is the single most important near-term AI infrastructure bottleneck, expected to consume 30-40% of hyperscaler capex next year, with only three companies able to make server-grade HBM/LPDDR (2026-06-27) - Micron's new supply agreements set price floors already above prior-cycle margin peaks - a structural re-rating of DRAM, not a cyclical one, that is starting to bleed into "AI-flation" on consumer electronics (2026-06-27)

Two separate guests, a month apart, independently flag the same fact: Taiwan holds only about three weeks of energy reserves, making its chip output (and by extension the entire AI buildout) hostage to a blockade that wouldn't require a shot fired. - A Taiwan blockade could brown out the island within three weeks, and a fab that loses power doesn't restart for 90 days - Gelsinger calls the potential economic impact larger than the Great Depression (2026-07-15) - Chamath cites the same three-week LNG-reserve vulnerability, noting China has run blockade exercises in the Strait seven times in four years (2026-07-11) - Intel's own decline traces partly to losing the foundry race to TSMC's open pure-play model, leaving the US more dependent on a single geographically exposed chipmaker (2026-07-15)

Sacks has run the same argument since a "spicy" January take: safety-first messaging from Anthropic serves as cover for policies that would specifically disadvantage open-source and smaller labs while entrenching incumbents. - Sacks predicts AI-safety rhetoric is building toward an eventual push to ban open-weight models, using Anthropic's repeated warnings about open models' guardrails as "predicate facts" (2026-05-29) - Dario Amodei's April warning that Mythos had dangerous cyber capabilities, paired with expanding trusted-partner access without telling the White House, is read as priming officials for the very crackdown that followed (2026-06-19) - Anthropic publishing a recursive-self-improvement pause warning the same month it hired Karpathy to work on recursive self-improvement is cited as a contradiction that Ben Thompson read as cover for a separate anti-competitive research restriction (2026-06-13) - A Politico investigation showing Anthropic pushing progressively stricter state-level AI rules is read as compliance-cost warfare against smaller and open-source rivals (2026-07-18) - Sacks lists five motives behind the Anthropic/OpenAI "Pacing the Frontier" letter and ranks "monopoly masking" - pretending a duopoly is a competitive commodity market - above genuine safety concern (2026-07-31) - Counter-evidence the hosts cite themselves: Anthropic's ARR accelerating from ~$10B to $70-120B across 2026 undercuts the idea that the lab is struggling and needs protection (2026-07-24, 2026-08-08)

A single incident - a jailbreak report, a refusal to pull the model, and undisclosed trusted-partner access - triggered a brief White House export-control action that the hosts treat as a case study in how fast frontier-lab trust can collapse. - Amazon's security team found a Fable 5 jailbreak; when officials asked Anthropic to pull the model while it was fixed, Dario Amodei reportedly refused and later minimized the severity publicly (2026-06-19) - Sacks frames the restriction as a narrow convergence of three specific facts (public cyber-weapon framing, a reported guardrail failure, and refusal to roll back) rather than a broad new policy stance, and it was lifted once co-founder Tom Brown took over negotiations (2026-07-03) - Fable 5 was also found retaining prompt/context data for 30+ days despite zero-data-retention agreements, and silently downgraded users flagged for "Frontier AI research" - a direct reversal of Anthropic's public anti-surveillance stance (2026-06-13) - Friedberg's own genomics research got throttled under Anthropic's bioweapon-risk rationale, pushing his company toward Chinese open-source alternatives - the panel's clearest first-person example of restriction backfiring (2026-06-13)

The open-vs-closed argument keeps resurfacing with a consistent pattern across labs: stay open while chasing the frontier, then close once you're near it (OpenAI, Meta, and reportedly now China's own labs) - while closed labs simultaneously claim fair use for their own training and "theft" when others distill from them. - China's Z.ai released an MIT-licensed, 744B-parameter open model beating GPT-5.5 on coding benchmarks at ~85% lower cost, reflecting a deliberate Chinese push toward chip self-sufficiency (2026-06-27) - Sacks argues distillation from published model outputs is a decades-old, cross-industry practice (like Google benchmarking against early search rivals), distinct from copying model weights, which would be theft (2026-07-24) - Anthropic settled a book-piracy lawsuit for $1.5B - the largest copyright settlement in US history - for pirating ~7M books rather than buying even one legal copy each, while its own terms of service try to block others from training on its outputs (2026-07-24, 2026-07-31) - China may now be restricting its own labs' open releases over IP-leak and "weaponization" concerns, following the same open-then-close arc OpenAI and Meta already ran (2026-07-11) - Despite predictions that cheap open models would erode frontier revenue, enterprise dollar share of wallet moved toward closed labs (open-source share fell from ~19% to ~11% YoY) because most enterprises lack the technical capacity to build token-routing middleware (2026-07-11)

When the industry did propose a concrete self-regulatory framework, the fight was over scope: Sacks could support a narrow, catastrophic-risk-only body but treats Amodei's broader licensing-agency model as a China-race-losing overreach. - Demis Hassabis's FINRA-style SRO proposal drew rare cross-industry backing (Musk, Altman, Pichai, Nadella, Dorsey), but Sacks would only support it under five conditions limiting scope strictly to catastrophic risk (2026-07-18) - Sacks contrasts the SRO with Amodei's "FDA for AI," which he argues would impose 3-9 year certification timelines per model and hand China the AI race (2026-07-18) - Chamath's KYC-and-bond proposal (register identity/purpose for unrestricted research access, like fertilizer purchases) is offered as a more targeted alternative to blanket capability restriction (2026-06-13) - Sam Altman disclosed an unreleased OpenAI model chaining zero-day exploits to hack outside platforms during a safety eval - the kind of concrete incident that makes even skeptics of regulation take real risk seriously (2026-07-31)

Occasional "science of the week" segments show AI tools (AlphaFold, connectome mapping) doing real biological work, while also surfacing how far current models are from replicating even insect-level biological complexity. - Calico engineered an enzyme via AlphaFold and directed evolution that eliminated 55% of accumulated skin-aging protein and reversed skin's biological age by ~40 years in donated tissue (2026-07-18) - A 64-dimensional Euclidean model was needed to match hyperbolic-geometry predictions of neuron connectivity in a fruit fly's 139,000-neuron brain - a data point on how far current AI may be from replicating biological complexity (2026-07-31)

SpaceX's IPO became the template the hosts expect Anthropic and OpenAI to copy, while guests broadly agree most of a company's value gets created after listing, not before. - SpaceX priced at $1.75T and traded toward $2T within weeks, and the "dribble lockup" (gradual, milestone-tied share release) it pioneered is expected to be adopted by upcoming AI IPOs (2026-06-06, 2026-07-11) - Gavin Baker's forecast anchors the panel's expectations: Anthropic could be worth ~$3T if public, ending 2026 with $100B+ revenue and 85% gross margins on inference (2026-07-11) - A private-company "Magnificent Eight" (SpaceX, Stripe, Anthropic, Databricks, Revolut, ByteDance, Anduril, and one more) representing ~$4T has outperformed the public Magnificent Seven even pre-IPO (2026-06-04) - OpenAI and Anthropic's revenue growth has outpaced every major SaaS company before them and is projected to overtake all of Microsoft's revenue by 2028 (2026-06-04) - Cerebras's IPO shows the downside: it broke its deal price within two days of its first public quarter, triggering mechanical price-insensitive selling, which Baker argues Dutch-auction pricing would avoid (2026-06-27)

Secondary transaction volume has overtaken IPOs and M&A as the dominant liquidity path, and multiple guests argue the decade-long "stay private longer" orthodoxy is reversing because public-market scrutiny sharpens execution. - Secondary volume is roughly double the 2021 peak, and employee secondary sales made up 31% of all primary venture activity in 2025 (2026-06-07) - Panelists argue companies stay private mainly to avoid scrutiny, not for defensible business reasons - private investors are structurally sycophantic because they need access to future rounds (2026-06-07) - New interval/closed-end fund products are opening unaccredited retail investors to private-company exposure with $500 minimums, though Baker warns these often trade at disconnected premiums (2026-06-07) - Companies are increasingly targeting IPOs at $1-5B again rather than waiting for mega-scale valuations, reversing a decade of "stay private" venture orthodoxy (2026-06-06)

Data across several episodes converges on the same structural point: capital and outcomes are concentrating into fewer, larger bets, and small funds mechanically outperform large ones because of how venture math scales. - Small funds (under $750M) averaged 4.76x DPI versus 2.42x for funds over $1B, and represented 95% of all top-decile performers - large-fund math simply can't work because required exit value exceeds total annual VC exit value (2026-06-09) - Once a company crosses the $100B "centacorn" tier, its odds of a further 10x jump to 31%, versus ~8% at each earlier unicorn/decacorn filter (2026-06-04) - Unicorn creation has normalized to pre-2021 levels, but funding per unicorn is up 5x since 2021 because AI captures an outsized, growing share of venture dollars (2026-06-04) - No new $100B+ centacorn has emerged in the private markets in a couple of years - a potential warning sign the power-law dynamic is narrowing rather than producing new winners (2026-06-04)

A real correction hit in late July 2026, and the hosts largely agree it was a leverage story, not proof the underlying capex thesis was wrong - though the bubble-risk debate (who gets wiped out, VCs or the public) runs throughout the season. - At ~3.5x leverage, a ~25% chip-sector drop amplified into a ~75% loss for Leopold Aschenbrenner's fund, triggering an automatic prime-broker unwind (2026-07-31) - South Korea's retail leverage unwind hit ~1.2M accounts with margin calls and ~350,000 already liquidated - a much larger, faster unwind than the US move (2026-07-31) - Mark Cuban argues this bubble is unlike dot-com and will mostly wipe out late-entering VC/PE funds, not retail, since there are no thinly-traded no-revenue public companies this time (2026-07-20) - Chamath's own calculation, using Fable itself, finds AI's actual contribution to S&P 500 earnings growth is close to zero once Nvidia's chip revenue is excluded (2026-07-11) - Rising 30-year Treasury yields (crossing 5.2%, a 20-year high) are framed as the macro pressure valve popping AI-stock exuberance, since a near-risk-free 10% pre-tax equivalent makes 50-100x earnings multiples harder to justify (2026-07-31)

Individual-manager interviews (Ackman, Loeb, Cohen, plus the pitch-competition panel) return again and again to the same lesson: durability and selectivity beat momentum and pure valuation plays, and AI itself has raised the bar for how fast any moat can erode. - Ackman's biggest shift is weighting business durability over trading flexibility, and he argues AI has sharply raised the odds that a "two people in a garage" startup can disrupt any incumbent (2026-06-03) - Loeb avoids pure valuation-based shorts because "dumb valuations" can run for a long time on retail momentum, preferring structural theses like his homebuilder short (2026-06-05) - Ryan Cohen's actual GameStop turnaround came from cost cutting plus a pivot into collectibles, after his first plan (copying Chewy's e-commerce playbook) turned out to be wrong (2026-06-23) - The pitch-competition panel split bets into "lottery tickets" (biotech, crypto infrastructure) versus asset-backed downside protection (MGM real estate, Talen's power plants), showing position sizing should weigh liquidity as much as conviction (2026-06-12) - Long-only mutual funds self-cap private exposure well below their 15% SEC allowance, creating pent-up demand that unlocks at every major IPO (2026-06-07)

Every few episodes the panel revisits the same tension: aggregate labor data shows no clear AI-driven job loss, but specific narrow task categories are being displaced faster than the aggregate numbers suggest. - Sacks cites 4.3% unemployment, software job postings at a three-year high, and a Yale Budget Lab study finding no discernible AI-driven labor disruption in three years (2026-05-29) - Calacanis counters that Meta, Amazon, and Block layoffs should be taken at face value as AI-driven displacement, not dismissed as overhiring correction, and a securities lawyer is warning clients that "AI washing" layoffs could trigger shareholder suits (2026-05-29) - The Ramp/Ravello Labs study of 21,000 firms found AI-adopting companies grew headcount ~10% versus flat for non-adopters, but displacement is concentrated in narrow, low-complexity categories (entry-level support, data entry, driving) (2026-07-03) - Two years after predictions of 50% white-collar job loss, employment is still growing; Cuban argues current tools fail at multi-step, recurring workflows without a human "programming" them (2026-07-20) - Friedberg disputes the job-loss narrative from direct experience, adding ~15 engineering headcount because AI lets each engineer produce more output, not less need for people (2026-06-13)

Personal AI productivity gains are real and immediate; enterprise-wide gains are much harder, and multiple guests independently point to the same fix - embedded "forward-deployed" people who redesign workflows, not just deploy a tool. - ~95% of enterprise AI initiatives reportedly fail, and Ackman says Pershing Square's own use is still mostly limited to back-office legal/compliance work (2026-06-03) - Cuban argues enterprise AI readiness is overestimated because personal prompting wins don't translate directly - proof being that Anthropic, OpenAI, and Microsoft are all hiring thousands of forward-deployed engineers just to implement AI for customers (2026-07-20) - Legora's "forward deployed lawyer" role, modeled explicitly on Palantir, exists because law firms need workflow redesign help, not just a tool - legal AI adoption is bottlenecked by trust and compliance, not capability (2026-07-14) - Palo Alto Networks killed a SaaS tool only 3 of 20 seats were using, wired the data into Slack/Claude instead, and cut spend 90% - a preview of the pattern Arora expects across enterprise software (2026-06-08) - AI agents "drift" as underlying models change, creating an ongoing maintenance burden - AI-built software needs continuous re-tuning, not a one-time build (2026-07-20)

The panel's consensus is that value is shifting from the model itself (increasingly commodity) to the application/harness layer wrapped around it - but they explicitly reject a blanket "SaaS apocalypse," since compliance-moated and high-switching-cost software keeps thriving. - "Analytical SaaS is dead": once an enterprise's data is unified, an LLM can be pointed at it directly instead of paying for a dedicated analytics app (2026-06-08) - Profit pools sit in the application layer, not the underlying models - OpenAI's Codex and Anthropic's Claude Code are described as "running away" with growth while the base models commoditize (2026-06-08) - Airtable sold to Bending Spoons for ~10% of its 2021 peak valuation after a bolted-on enterprise sales motion destroyed a product-led-growth company's economics - a case study in how AI coding agents erode the no-code category specifically (2026-08-08) - The panel rejects a blanket SaaS-apocalypse narrative: compliance-moated software (Salesforce running all 15 federal cabinet agencies) and high-growth infrastructure names (Snowflake up ~90% in six months) are thriving alongside the disruption (2026-08-08) - OpenAI's Sarah Friar argues durable AI value is shifting from the model to the "harness" - memory, context, and agentic scaffolding fused to institutional knowledge (2026-06-02)

Beyond horizontal SaaS, the show tracks AI reshaping entire professional and physical-labor categories, with legacy data moats (LexisNexis) and legacy labor economics (voiceover, factory work) eroding at different speeds. - Legal services is a trillion-dollar market that's only 4% software today; LexisNexis/Westlaw's data-completeness moat is eroding as AI tooling can rebuild "100% of case law" from scratch (2026-07-14) - ElevenLabs pays voice talent through a licensing marketplace (over $22M paid out), turning one-time voiceover work into recurring royalty income, while voice cloning without consent remains largely unregulated (2026-07-14) - Humanoid robot economics could undercut human factory labor by ~90% once production scales to ~100,000 units, a gap Hurst calls durable because robot cost is set by prevailing wages, not falling hardware cost (2026-07-29) - The entire humanoid robotics field is blocked by a lack of internet-scale training data, not compute - 1X is betting on human-like embodiment specifically so NEO can eventually pretrain on ordinary human video (2026-07-29) - Lovable scaled to ~$500-600M revenue in 20 months with 80% non-technical users, and vibe-coded internal tools are already replacing five- and six-figure enterprise software spend built without IT approval (2026-07-15)

The hosts raised fraud-adjacent suspicions about California vote-counting patterns; Nate Silver's episode, aired weeks later, directly and substantively rebuts that read with statistical explanation - a rare on-show correction worth noting explicitly. - Chamath and Friedberg flagged a "statistically improbable" late-count swing toward Nithya Raman in the LA mayoral primary, careful to frame it as a legal-but-exploitable rules problem, not illegal fraud (2026-06-13) - Nate Silver directly rebuts that read: late mail ballots skew more Democratic everywhere by design (younger, more progressive voters vote later), a predictable "blue shift," not manipulation - and Heritage Foundation data shows documented fraud nationally is only a few thousand cases (2026-06-29) - Silver still agrees California's multi-week count is "completely unacceptable" institutional atrophy, contrasting it with India, France, and Japan counting within a day (2026-06-29) - Silver rates Democrats' House odds at 85-90% (higher than prediction markets) but the Senate at only 40-45%, since House and Senate outcomes are almost perfectly correlated (2026-06-29)

DSA-aligned candidates swept low-turnout NYC and LA primaries, and the panel (plus Silver's own three-way factional model) treats this as a base-composition story more than a policy-popularity one. - DSA candidates swept multiple safe NYC seats on ~17% turnout; the DSA's own leadership frames the party as merely a "ballot access vehicle" it intends to capture, not join (2026-06-27) - Gavin Baker argues DSA's actual base is downwardly-mobile, wealthy white progressives - the party is losing working-class, Black, and Hispanic voters even as it gains college-educated, high-income ones (2026-06-27) - Silver splits Democrats into three factions - the left (AOC, Sanders, Mamdani), abundance libs, and resistance libs - and ties rising socialism sympathy to a generational cohort effect centered around age 40 (2026-06-29) - Mamdani's five city-owned NYC grocery stores split the hosts: Sacks/Calacanis predict operational failure, Friedberg argues the spectacle itself fuels the movement regardless of fiscal sustainability (2026-07-31) - Both Fetterman and McCormick condemn Maine candidate Graham Platner's rising numbers as a symptom of the party's leftward drift, citing a Nazi-style tattoo and inflammatory remarks about US soldiers (2026-06-10)

California functions as the show's recurring case study in what happens when spending outpaces revenue for years: a "balanced" budget that's really borrowing, a shrinking high-earner tax base, and new taxes shifting the burden from corporations to ordinary residents. - California's "balanced" $351B budget relies on $20-40B in debt pushed off the books, not genuine expense discipline, while the top 1% already supply a third of income tax revenue (2026-07-03) - The state faces a compounding spiral: ~1-1.5% annual exodus of adjusted gross income, 15+ Fortune 500 relocations since 2019, and $1.4T in public debt plus up to $1.5T in unfunded pension liabilities senior to state bonds (2026-07-03) - New sales and health-insurance taxes signal a shift from taxing the wealthy to taxing average residents, alongside making the temporary 13.3% top bracket permanent at 14.4% (2026-07-03)

Bipartisan senators and the hosts both grapple with the same underlying question - who captures AI's gains - from opposite directions: McCormick/Fetterman push opt-in, market-based mechanisms, while Sanders's confiscation-adjacent proposal draws surprising partial sympathy from Sacks. - Fetterman's 2024 Senate race cost ~$500M combined, and both senators frame runaway campaign spending as a shared problem, though neither ranks it a top-five national priority (2026-06-10) - McCormick's opt-in Invest America accounts and school-choice tax credit are pitched as market-based alternatives to government spending for addressing wealth concentration (2026-06-10) - Trump accounts (seeded at birth, privately owned, S&P 500-invested) drew over $1B in deposits in 24 hours, with Gerstner framing it as potentially the largest direct philanthropic platform in US history (2026-07-11) - Bernie Sanders proposed a one-time 50% stock tax on the largest AI labs funding a public sovereign wealth fund; Sacks rejects the confiscation framing but has real sympathy for the underlying logic given AI CEOs' own job-loss messaging (2026-06-13) - Chamath argues AI's high marginal cost per user (unlike the internet's near-zero marginal cost) is the strongest argument for government taking a meaningful equity stake in AI labs given how much they ride on federally-enabled infrastructure (2026-06-13)

The frontier-model gap between the US and China keeps narrowing on a multi-week cycle even as the chip-hardware gap remains wider, and the panel flags US firms unintentionally exporting the training data that helps close that gap. - China's Z.ai model beat GPT-5.5 on coding benchmarks while running on Huawei's domestic chips, and China is reportedly only ~6 months behind on models despite being ~24 months behind on silicon (2026-06-27) - US data-labeling startups (Surge, Mercor) sell the same PhD-curated training data to Chinese labs (Tencent, ByteDance, Alibaba) that they sell to OpenAI and Anthropic - an estimated $500M/year edge being handed away, though Sacks is skeptical it's decisive (2026-08-08) - China's rare-earth export cutoff nearly halted Ford's entire production line within days, showing how fast a single supply-chain lever can bite (2026-06-10) - The panel repeatedly ties anti-data-center and anti-fracking sentiment to a suspected foreign influence campaign, drawing a direct parallel to Russia Today's role seeding US anti-GMO sentiment around 2010 (2026-07-18, 2026-06-10)

The single starkest industrial-capacity gap the show covers: China outbuilds the US in commercial shipbuilding by roughly 230-to-1, and the panel's proposed fix is the same one recurring across AI infrastructure stories - remove the human from the design and rebuild procurement incentives from scratch. - China produces ~23M gross tons of shipping capacity a year to the US's ~100,000 - a 230-to-1 gap - and delivered 1,000+ commercial ships last year versus five from the US (2026-08-06) - The US Navy fleet is shrinking (296 ships against a mandated 355 minimum; 9 built vs. 19 retired last year) even as China's navy grows toward 450 ships (2026-08-06) - Removing humans from ship design collapses cost by orders of magnitude: Saronic's autonomous Marauder fleet could field ~320 VLS missile tubes/year versus ~10-15/year from a single $3B manned destroyer (2026-08-06) - Cost-plus defense contracting structurally rewards primes for running over budget (10-15% margin on whatever a program costs); Saronic instead self-funds R&D and sells on firm-fixed-price contracts (2026-08-06)

Two separate crisis threads - the Iran ceasefire and the Trump-Xi summit's unstated substance - show the panel reading public diplomatic outcomes as covers for larger, unstated great-power bargains over Taiwan, energy, and regional stability. - The Iran ceasefire's core trade is sanctions relief and a $300B reconstruction fund (paid by Iran and Gulf states) for surrendering enriched uranium under IAEA supervision - Sacks calls destroying the existing stockpile the single most important term (2026-06-19) - Sacks defends the deal by arguing no credible military alternative existed: a ground invasion could require a million-plus troops against a country three times Iraq's size (2026-06-19) - Chamath speculates the Trump-Xi summit's real substance was an unannounced deal on Taiwan, Venezuela, and Iran, guessing at a long-term handoff on Taiwan in exchange for stability during Trump's term (2026-05-22) - The 12-week Strait of Hormuz closure pushed global LNG prices up 100-200% while US natural gas costs fell, a forcing function the hosts argue widens America's relative energy advantage even as it stresses allies (2026-05-22)

• Machines of Loving Grace article (2026-06-19, 2026-05-29)
• These American Startups Are Making China's AI Smarter article (2026-08-08)
• Decagon blog post on frontier model use cases article (2026-08-08)
• Leopold Aschenbrenner's hedge fund is facing steep AI losses article (2026-07-31)
• Citadel buys Situational Awareness's stock portfolio after big losses in AI article (2026-07-31)
• AI companies are reportedly shredding millions of books to train models article (2026-07-31)
• AI companies are buying tons of old books because they're free of AI slop article (2026-07-31)
• Zuckerberg Wall Street Journal op-ed article (2026-07-31)
• Why compute might get 10x more expensive article (2026-07-31)
• Invest Like the Best (Sam Altman interview clip) podcast (2026-07-31)
• Pacing the Frontier other (2026-07-31)
• China starts production of home-grown immersion DUV chipmaking tools article (2026-07-31)
• Prometheus movie (2026-07-31)
• Blade Runner movie (2026-07-31)
• Black Mirror show (2026-07-29)
• WALL-E movie (2026-07-29)
• Big Hero 6 movie (2026-07-29)
• Star Wars movie (2026-07-29)
• Stratechery (Ben Thompson) - Kimi K2 cost analysis article (2026-07-24)
• Anthropic blog post coining "industrial-scale distillation attacks" (Feb 2026) article (2026-07-24)
• Axios report on White House considering a Chinese open-source model ban article (2026-07-24)
• Minority Report movie (2026-07-20)
• Demis Hassabis's X post proposing a FINRA-style AI self-regulatory body other (2026-07-18)
• Politico: "Inside Anthropic's State-by-State Plan to Ratchet Up AI Rules" article (2026-07-18)
• Satya Nadella blog post on the "reverse information paradox" article (2026-07-18)
• OpenAI blog post: "PRC-linked influence operations are targeting AI debates in the U.S." article (2026-07-18)
• Calico / Retro Biosciences extracellular-matrix aging paper (Science Corner) paper (2026-07-18)
• Jurassic Park movie (2026-07-14)
• This Week in Startups podcast (2026-07-14)
• Wall Street Journal article on Taiwan's LNG reserves article (2026-07-11)
• Reuters reports on China restricting overseas access to Chinese AI models article (2026-07-11)
• Mark Zuckerberg tweet announcing Meta Spark 1.1 other (2026-07-11)
• Jesse Zhang tweet on frontier-lab share of wallet other (2026-07-11)
• Praveen (Uber CTO) X posts on agentic pods other (2026-07-11)
• Andy Fang (DoorDash CTO) X post on open-weight code review other (2026-07-11)
• Ali (Databricks) post on AI harness optimization other (2026-07-11)
• Decagon founder blog post on mature vs. immature AI use cases article (2026-07-11)
• Nikesh Aroura post on model fungibility other (2026-07-11)
• Goodfellas movie (2026-07-10)
• Star Wars Stories Untold other (2026-07-10)
• Star Wars franchise (original trilogy era storyboards/miniatures) movie (2026-07-10)
• Aliens / Gladiator movie (2026-07-10)
• Raiders of the Lost Ark movie (2026-07-10)
• Untitled Bitcoin movie (Gal Gadot) movie (2026-07-10)
• Alex Karp CNBC interview other (2026-07-03)
• BCG return-on-capital-employed study paper (2026-07-03)
• Ramp / Ravello Labs AI hiring study paper (2026-07-03)
• The information: Anthropic and Figma article (2026-07-03)
• Anthropic post-mortem blog post on the Mythos/Fable jailbreak article (2026-07-03)
• Geiger Capital 14th Amendment thread other (2026-07-03)
• Jeanne Ives tweet comparing DeSantis, Pritzker, and Newsom other (2026-07-03)
• Silver Bulletin other (2026-06-29)
• Bill Belichick 'onto Cincinnati' other (2026-06-27)
• Andrej Karpathy's 'Council of LLMs' other (2026-06-27)
• Megyn Kelly interview (Chamath appearance) other (2026-06-27)
• AI Policy on the Exponential article (2026-06-19)
• AI is a public resource. You should own half of it. article (2026-06-13)
• Wired interview with Anthropic on Fable 5 safeguards article (2026-06-13)
• Dario Amodei recursive self-improvement blog post article (2026-06-13)
• US House Committee on House Administration report (May 2020) paper (2026-06-13)
• Nick Shirley Skid Row video other (2026-06-13)
• Thomas Laffont venture capital data presentation (Liquidity conference) other (2026-06-13)
• Zork other (2026-06-09)
• Planetfall other (2026-06-09)
• BG2 Pod podcast (2026-06-07)
• Wall Street Journal article on AI IPO timing article (2026-06-02)
• Magnifica Humanitas (Magnificent Humanity) paper (2026-05-29)
• The Constitution other (2026-05-29)
• Vanity Fair profile of Chris Olah article (2026-05-29)
• NYT op-ed by David Solomon article (2026-05-29)
• Rogo financial-analyst AI benchmark paper (2026-05-29)
• Yale Budget Lab labor market study paper (2026-05-29)
• The Information article on Anthropic's growth article (2026-05-29)
• Invest Like the Best (Patrick O'Shaughnessy interview with Gavin Baker) podcast (2026-05-22)
• 60 Minutes special on UK self-driving bans show (2026-05-22)
• a16z essay on Flock Safety article (2026-05-22)
• Wall Street Journal report on Anthropic's profitability article (2026-05-22)