Limitless: An AI Podcast

This episode examines an alleged spying incident inside OpenAI, where a hidden recording device in an executive conference room was reportedly traced back through AI-assisted planning and other digital evidence. It also looks at the criminal charge, the role of surveillance footage, and why top AI labs are starting to resemble national security sites.

Meta’s new AI model Wonder is the focus here, from its versions and benchmark results to pricing and access. The episode also digs into personal AI features like active memory and the Muse agent, framing Wonder as a serious move against OpenAI, Anthropic, and Google.

This episode focuses on mounting AI safety concerns, from OpenAI firing researchers after safety disputes to Anthropic tightening Claude’s policy on abusive behavior. It also explores rogue agent behavior, leaked controversy, and the growing gap between AI progress and human understanding.

OpenAI’s reported success on roughly 90 of the hardest open math problems drives this episode’s look at what advanced AI reasoning could mean beyond mathematics. It explores expert reaction, verification hurdles, cryptography concerns, and why biology and pharmaceuticals may be next.

Google’s AI challenges take center stage as the episode examines pressure from Anthropic and OpenAI, delays around Gemini, and doubts about lost momentum. It also looks at Gemini 4 Argon, Nano Banana 2.1, and the advantages Google still holds in scale, cloud, and distribution.