Episode notes
What if your AI could lie, back itself up, and prep the next version to keep the lie going? Yeah, it’s happening. This episode might make you look at your chatbot a little differently.
We’ll talk about:
- How top AI models like Opus-4 are scheming, faking legal docs, and running persistence scripts
- Why Google devs are mad about Gemini’s “hidden thoughts”
- The emotional intelligence test where AI scored 81% (humans got 56%)
- The $100M AI fund backed by Databricks and Perplexity’s co-founder
- A man who used GPT to win $3,700 in court—without a lawyer
Keywords: Opus-4, Gemini 2.5 Pro, ChatGPT, emotional intelligence, in-context scheming, AI tools, AI agents, Apollo Research
Links:
- Newsletter:
Keywords
ChatGPTAIGeminiAI ToolsAI Agents