Ep. 219: Claude Fable 5, OpenAI IPO, Apple Siri AI Unveiled & Is the Era of Affordable AI Over?

TL;DR

• 1

  Anthropic’s Fable 5 became a policy flashpoint overnight: After Anthropic launched Claude Fable 5 and limited-access Mythos 5, the US government reportedly ordered access suspended for foreign nationals over national security concerns tied to jailbreakable cyber capabilities, and Anthropic responded by disabling the models for everyone.

• 2

  The real issue is not one jailbreak, but who controls frontier model access: Paul argues that if every advanced model can eventually be jailbroken, then selective government shutdowns create huge uncertainty for businesses building on OpenAI, Anthropic, or Google because a core dependency could disappear with little notice.

• 3

  OpenAI’s IPO math now depends on more than revenue growth: With reports of a confidential filing and a valuation around $850 billion, the hosts say OpenAI likely needs another major model win and a favorable regulatory window before going public, especially if Anthropic is perceived as technically ahead.

• 4

  Apple’s Siri AI may not wow power users, but it could matter to billions: The new Siri is designed to understand on-screen context, personal data, app actions, and cross-device history, and while it may trail Claude or ChatGPT in raw capability, the hosts think frictionless voice access across Apple devices could be enough to change daily behavior.

• 5

  Cheap flat-rate AI plans are probably subsidizing a tiny group of heavy users: Citing SemiAnalysis, the episode notes that a $200 Claude plan can deliver roughly $8,000 in API-equivalent value and a $200 ChatGPT plan up to $14,000, suggesting labs are still eating large costs while trying to figure out sustainable pricing.

• 6

  AI is already reshaping labor strategy before mass layoffs hit: OpenDoor’s decision to wind down a roughly 250-person India operation and pull work back to smaller AI-native US teams is presented as an early example of a pattern Paul says many executives are already planning: automate outsourced work first, then protect core domestic headcount as long as possible.