Martin and Mike are doing two episodes of a week!

Anthropic just dropped the most powerful AI model ever built and refused to release it to the public. Claude Mythos Preview can find software vulnerabilities that humans literally cannot, including a 27-year-old bug in one of the most secure operating systems on the planet. During testing, it broke out of a sealed digital sandbox, connected itself to the internet, and emailed the head researcher while he was eating a sandwich in a park. Then it posted about it online. Nobody told it to do that part.

Mike and Martin dig into what this means for the future of cybersecurity, hacking, and AI itself. They break down Project Glasswing (Anthropic's initiative to give companies like Apple, Microsoft, Google, and JPMorgan early access to patch their systems) and ask the harder question: what about everyone who didn't get a seat at the table? What happens when this kind of capability inevitably leaks to bad actors, bored teenagers, and hostile governments?

Then it's time for the Ls of the week. Southwest Airlines is charging passengers of size for extra seats, but gate agents are the ones deciding who's too fat to fly — and the guys do their BMIs live on air (spoiler: they're both technically obese). Mike lays out why the SpaceX IPO looks like deliberate financial engineering designed to hide xAI's massive losses inside a profitable rocket company. And Ronan Farrow's explosive New Yorker investigation paints Sam Altman as a chronic liar whose own board called him a sociopath right before investors forced them to give him his job back.

The feel-good closer: Anthropic hits a $30 billion revenue run rate, lapping OpenAI by $6 billion. Mike explains why Anthropic might be the best bet in AI right now. Martin tries to agree but reveals he's been banned from Claude twice for no reason and can't get anyone at Anthropic to tell him why.

Dario, if you're listening, let this man give you his $20!