🎙️ EP 96: Ex-OpenAI CTO Just Fixed a Bug Google Couldn't for 5 Years

One tiny team just did what OpenAI, Google, and Meta couldn't, fix the most frustrating bug in AI. Mira Murati’s new startup dropped open-source code that could finally make AI stable. That’s just the start.

We’ll talk about:

• The shocking AI bug that made ChatGPT give different answers to the same question
• How a $12B startup fixed it before anyone else (and why it matters)
• A side-by-side test of ChatGPT vs. DeepSeek in healthcare, the results were wild
• A look into what’s really coming next from Thinky Labs (spoiler: it’s not just research)

Keywords: Mira Murati, ChatGPT, DeepSeek, Thinking Machines Lab, LLM, reproducibility, AI tools, batch invariance, GPT-4o

Links:

1. Newsletter: