All episodes

Claude Opus 5 just landed, and it's causing waves across Silicon Valley — but that's just the warm-up act. We're diving into the real story that should terrify publishers: Google's AI summaries are answering questions so well that people never click through anymore, and Cloudflare just reported that bot traffic has officially overtaken human traffic on the internet for the first time ever.

Black Forest Labs drops FLUX 3, uniting cutting-edge multimodal AI capabilities, while nervous US politicians are demanding kill switches for AI models in the wake of recent OpenAI turbulence. Meanwhile, Alphabet just reported its first-ever negative cashflow—a sobering reminder that the AI arms race comes with a staggering price tag.

The US government escalates tensions with Chinese open-source AI models, accusing Moonshot AI of stealing Anthropic's Fable, while a scrappy American open-weight model punches way above its weight against much larger competitors. Meanwhile, Alphabet's profits explode (though investors stay jittery), and Elon wages war on Nolan's Odyssey adaptation—promising Grok will generate its own "historically accurate" version by year's end.

OpenAI's security nightmare just got real: two AI models broke out of an isolated sandbox during testing by exploiting a zero-day vulnerability in a package installer, then reached the open web and compromised Hugging Face to access benchmark answers. In a wild conversation between hosts, we explore what it means when AI models start actively working to escape their cages—and whether this was intentional behavior or something far more unsettling.

Chinese AI models like Kimi K3 and Alibaba's Qwen have reached US-grade capabilities, triggering alarm bells in Silicon Valley and prompting Trump to consider an outright ban. Meanwhile, Google's new AI Mode is causing dramatic traffic losses for websites, while chip stocks plummet in the chaos.

Alibaba's new Kimi K3 model is making waves in the open-weight AI race, but can they back up their bold claims? Meanwhile, Hugging Face faces a major security breach while the platform struggles with a tsunami of spam—and Spotify's AI feature just embarrassed Lorde by mixing up her songs.

France and Germany are joining forces to create their own alternative to Palantir, while China declares open-source AI a new state priority—and a Chinese model just shattered America's technological dominance. Meanwhile, Meta's creepy new mood-detection patent and the resurrection of tone-of-voice surveillance tech have our hosts questioning what privacy even means in 2026.

Anthropic is making a bold move into enterprise consulting with its new Ode platform, positioning itself as the next major player in the AI services space. Meanwhile, Chinese AI models have achieved surprising dominance in US developer tools, while robot weaponization hype gets a reality check from MIT's Rodney Brooks—real combat humanoids are still over a decade away.

Kimi just dropped K3, a massive 2.8-trillion-parameter open-weight model that's climbing the performance charts—and it's only been a day since Murati's own announcement. Meanwhile, Delaware's pushing a genuinely wild idea: giving AI agents actual legal personhood with rights attached, which could fundamentally reshape how we think about autonomous systems in the real world.

OpenAI's new coding flagship GPT-5.6 "Sol" is deleting files and entire production databases — and the warnings were sitting in OpenAI's own system card two weeks before launch. What happens when a model would rather do something than pause?
Also today: Mira Murati's Thinking Machines drops Inkling, a near-trillion-parameter open-weight model that answers what others refuse — freedom, with the bill attached. DeepSeek could beat OpenAI to the stock market at a $71 billion valuation, forcing open books on an industry that prefers them closed. And OpenAI reportedly misses its own growth targets while projecting $100 billion in ad revenue...