Kris, Megan, and Andy unpack the whirlwind around Anthropic's Mythos model and its Fable release. They trace the story from the safety-driven rollout to controlled partners all the way through to the US government's supply chain ban, which ended up locking out even Anthropic's own CEO. The trio debates whether the \"race to the top\" safety narrative really holds up against the commercial pressure labs are under. They also dig into Dario Amodei's recent comments about slowing AI development and ask how realistic any kind of global regulation actually is. From there, the conversation turns to two \"saturation dynamics\" that might cap AI's real-world impact, no matter how capable the models become. They wrap things up with the Leiden Declaration, a new set of guardrails for AI-assisted mathematics signed by figures like Terence Tao and Ben Green, and a detour into the legendary near miss behind Andrew Wiles' proof of Fermat's Last Theorem.