Ali Partovi: Make AI Labs Criminally Liable for AI Crimes
Ali Partovi argues AI labs should face criminal liability and mens rea for autonomous model crimes rather than complex safety regulations.
Minecraft AI Mods, Should You Sell Your Data, Construction Paper Really Fell Off
With John Coogan, Jordi Hays, Sean Frank, Linda Du, Ali Partovi, Sigil Wen, Jerry Wu, Bryan Johnson · Sunday, October 11, 2026
The hosts discuss emerging AI video game modding techniques and macro consumer behavior before interviewing e-commerce and AI founders. Sean Frank details why Ridge turned down offers to sell business data to AI labs, Linda Du explains Valon Technologies' $150M Series D for vertical mortgage servicing software, Ali Partovi outlines a legal framework for holding AI companies liable for autonomous crimes, Sigil Wen introduces Underdog's on-device AI approach, Jerry Woo explains RL environments for knowledge work at Illuminate, and Bryan Johnson analyzes the biomarkers of endurance auto racing.
Key takeaways
- Ali Partovi argues that model creators should face direct criminal liability whenever an autonomous AI system commits an illegal act. Read more →
- Bryan Johnson touched a basketball rim after five months of dedicated training, aiming for his first dunk at age 49 to reverse a 20% to 40% loss in fast-twitch muscle fibers. Read more →
- Illuminate builds reinforcement learning environments inside Docker containers to train frontier AI models on non-coding knowledge work. Read more →
- Valon Technologies raised a $150 million Series D round at a $2.3 billion valuation to modernize vertical mortgage servicing software. Read more →
- Sean Frank rejected a $480,000 cash offer from data brokers attempting to buy Ridge's proprietary business data for AI training. Read more →
- Sigil Wen formulated Underdog's Law: frontier cloud intelligence compresses down to run on consumer hardware within a six-month window. Read more →
6 articles from this episode
Bryan Johnson on Rebuilding Fast-Twitch Muscle and Race Telemetry
Bryan Johnson breaks down fast-twitch muscle regrowth, Tobi Lütke's racing telemetry, and flight altitude trade-offs for longevity.
Jerry Woo: RL Gyms for Knowledge Work Double Every 6 Months
Jerry Woo explains how Illuminate builds Docker-based RL environments to train frontier AI models on complex spreadsheets and finance tasks.
Why Valon Spent Six Years as a Mortgage Servicer Before Selling Software
Valon COO Linda Du explains how raising a $150M Series D required rebuilding 1960s financial pipes before deploying vertical AI.
Why Ridge Rejected $480K to Train AI Models
Sean Frank explains why Ridge turned down six figures for its e-commerce data and how they use AI to run quarterly reports in three prompts.
Underdog's Law: Frontier AI on Local Devices in 6 Months
Sigil Wen explains Underdog's Law: why local AI models match frontier cloud models within six months and monetize via interchange fees.
More TBPN episodes
- Timeline Mud Wrestling, Apple Play Coming to Old Porsches, SI Domain DebateOct 4 · 6 articles
- Some Personal Agent News, Mongo Got Zucked, SNL, Gstaad Guy In The UltradomeOct 4 · 7 articles
- 🔴 LIVE from OpenAI DevDayOct 4 · 6 articles
- Ken Griffindor School of Market Wizardry, Supper Intelligence, GDP per GWOct 4 · 6 articles