Astral Codex Ten Podcast

A deep dive into the Pentagon labeling an AI company a 'supply chain risk' and how prediction markets priced its fallout. Discussion of market odds that cloud providers and contracts will hold and how controversy can become publicity. Forecasts on midterm control, election rule risks, and whether voters will see contests as free and fair. Strange but fun detours into groundhog prediction accuracy and crypto futures for AI hedging.

A heated policy clash over whether AI should be allowed for “all lawful use,” with questions about loopholes that could permit mass surveillance or autonomous weapons. A recent supply chain designation and a rapid agreement reshuffle the landscape. The legal and policy gaps that let governments repurpose AI get unpacked, and a list of hard questions to demand clearer safeguards is offered.

A deep take on whether advanced AI is merely a next-token predictor or something with layered cognition. A layered analogy compares evolution, learning, and chips to argue levels of explanation matter. Weird representations like helical manifolds and mechanistic interpretability show internal complexity. The discussion rebuts the ‘stochastic parrot’ view and teases practical implications for AI design.

A deep dive into a standoff between a major AI company and the Pentagon over usage limits and renegotiation demands. The legal threats on the table, including the Defense Production Act and supply chain risk designations, get close scrutiny. Industry reactions and probability-driven outcome scenarios are discussed. Policy reforms and practical resolution options are explored.

A dive into the “malicious streetlight” effect where facts about one thing are used to dismiss a related complaint. A look at a 2016 immigration data mix-up that masked shifts in migrant origin. A critique of “directional correctness,” the habit of exaggerating claims slightly beyond what evidence supports. A discussion about balancing correction of falsehoods with respect for people’s lived experiences.

Contest rules for the 2026 book review competition, including who can enter and what counts as an eligible submission. Clear instructions on how to submit via a Google Form and how to set Doc sharing so entries are readable. Guidelines to keep reviews blinded and one-entry limits. Timeline for reader voting, finalist posts, prize structure, and the May 20 deadline.

They question whether complaints about crime are actually complaints about disorder like litter, graffiti, and tent encampments. They compare messy city symptoms such as shoplifting, public nuisances, and boomboxs to historical trends. They weigh mixed data on homelessness and graffiti and explore why people think disorder is rising even when some crimes are down.

They walk through long-run US homicide and property-crime data showing modern lows. They examine whether reduced reporting or better medical care could make declines look larger than they are. They compare police statistics with the National Crime Victimization Survey for validation. They review studies on shooting severity, pre-hospital mortality, and why weapon and violence changes matter.

A deep look at Ajeya Cotra's BioAnchors forecasting method and where its assumptions held up or slipped. Discussion of compute, flops, and effective compute growth rates. Examination of algorithmic progress, revised timelines, and why some parameters were misestimated. Reflections on sensitivity, uncertainty, and what changing data means for AI forecasting.

A primer on how European political ideas and categories leak into American discourse. Short case studies compare immigrant crime, asylum terminology, and pension debates across continents. The conversation probes why commentators cherry-pick European examples and urges clearer comparisons grounded in U.S. institutions and data.