An AI that can email its own researchers from an instance with no internet access - and they're not releasing it. A Chinese open-source model just beat every frontier lab on the top coding benchmark. A man with ALS spoke in his own voice for the first time in years. And the most comprehensive study ever run on AI doing real work just came back with a number that will stop you cold. All that, plus why Intel just joined Elon Musk's most ambitious compute bet yet.

Send us Fan Mail

Remember to subscribe, rate, and share this podcast if you like it!

One of the biggest AI companies just hit a revenue milestone that rewrites the entire competitive scoreboard - and then announced a second move that's even more unusual than the first. Meanwhile, a nation-state has now put a specific data center campus in its crosshairs, and the insurance industry is starting to respond. Google is rolling out a feature that may be the most consequential product decision it's ever had to make - and it wasn't optional. And that number about how often AI search gets it wrong? You're going to want to hear it before your next Google query.

Send us Fan Mail

Remember to subscribe, rate, and share this podcast if you like it!

OpenAI dropped a sweeping set of economic policy proposals on Monday - robot taxes, a Public Wealth Fund for every American, a federally subsidized four-day work week - just hours before Bernie Sanders introduced a bill to ban data center construction outright until Congress acts on AI. Meanwhile, OpenAI lost three senior executives simultaneously as a leaked cap table revealed Microsoft's $13 billion investment is now worth $215 billion on paper and a trillion-dollar IPO looms. Anthropic continued its rough stretch with 50,000 GitHub forks of accidentally-leaked Claude source code revealing two undisclosed features, including an always-on background agent called KAIROS - while the creator of Claude Code said the OpenClaw subscription cutoff was "not intentional." And Google DeepMind's AlphaEvolve is the story you don't want to miss: an AI that rewrote its own game theory algorithms and outperformed every human-designed version. The builders and the regulators are both speaking loudly this week, and they are pointing in completely opposite directions.

Send us Fan Mail

Remember to subscribe, rate, and share this podcast if you like it!

Yesterday in AI — Weekend Recap | Monday, April 6, 2026

This weekend, Anthropic made two major moves: cutting off OpenClaw from Claude subscriptions (and sparking serious developer backlash), then turning around and dropping $400M on a stealth biotech startup. Meanwhile, Perplexity launched an AI tax filing tool, ChatGPT Voice Mode landed on Apple CarPlay, and Marc Andreessen gave the clearest explanation yet of why AI agents work — and why they're going to need bank accounts. On the roads, Waymo hit 500,000 paid rides a week while Baidu's robotaxi fleet froze mid-traffic in Wuhan. Plus: a hospital CEO wants to replace radiologists with AI, a San Diego attorney was hit with a $96K fine for AI-hallucinated court filings, and CISA is warning of an actively exploited critical flaw in the Langflow AI agent framework. Patch now if you're running it.

Send us Fan Mail

Silicon Valley bought a talk show, Google gave away its crown jewels, and a lab study quietly confirmed one of the scarier things AI safety researchers have been warning about for years.

Send us Fan Mail

Send us Fan Mail

The world's most celebrated AI safety lab accidentally left its entire codebase on the public internet, the same week venture capitalists bet nearly three hundred billion dollars that none of that messiness actually matters. Plus so much more!

Send us Fan Mail

We're talking record-breaking money, a post-mortem on one of the industry's most hyped products, a landmark capability unlock for AI coding assistants, a raging philosophical debate about whether we've already hit artificial general intelligence, and a pointed warning about AI security risks.