Issue 345: The AI we dream about

Getting our AI agents to act reliably and predictably can feel like a pipe dream. But who are we if not dreamers? In this issue, we’re dreaming—not hallucinating—up ways to make our AIs do what we want, whenever we want.

This week, we’re chatting with the creator of Lean, Leo de Moura, about proving correctness in AI and how automated reasoning complements probabilistic models. We sit down with LinkedIn’s Praveen Bodigutla for a conversation on their new hiring assistant and the four-layer cognitive memory system powering it. Plus, we explore why responsibly designed dev workflows are the key to seeing—and defeating—shadow AI attacks.

Folks from around the web are dreaming just like we are. And some of those dreams become reality—like when one dev this week broke Auto Mode in Opus 5. Sometimes those dreams remain impossible to achieve—like actually being able to see a black hole. But there’s no use in losing sleep over any of it because we’re making all of your knowledge day dreams come true…all you need to do is head to the links below.