The Artificial Intelligence Show #207: OpenAI vs. Anthropic Feud, Claude Mythos Leak, Brutally Honest CEOs & Data Center Moratorium
187 snips
Mar 31, 2026 A long-running AI rivalry turns deeply personal, with broken promises, power struggles, and huge stakes for business and geopolitics. There is also buzz around a leaked next-gen model, blunt CEO talk about job disruption, political battles over data centers and deregulation, rising agent security nightmares, and Apple’s surprising Siri reset.
AI Snips
Chapters
Transcript
Episode notes
Mythos Leak Signals Another Capability Leap
- Anthropic’s leaked Mythos materials suggest a near-term jump in model capability, especially in reasoning, coding, and cybersecurity.
- Paul Roetzer says labs always see six to twelve months ahead, so current user experience badly understates what is already trained and being safety-tested.
OpenAI Stages Suggest Faster Than Expected Takeoff
- Paul Roetzer argues AI progress is moving from chatbots and reasoners toward agents and innovators far faster than most businesses expected.
- He says OpenAI’s 2024 internal ladder went from level one to emerging level four in roughly twenty months, with early level five signs possible in some industries.
CEOs Are Finally Saying The Quiet Part Out Loud
- Executives are starting to say publicly what many have said privately for months: AI will remove large amounts of work and workers cannot simply opt out.
- Paul Roetzer thinks AI-forward managers stay valuable, but entry-level roles face the hardest squeeze because prompting systems can absorb much of the tactical learning work.