All-In with Chamath, Jason, Sacks & Friedberg · Tuesday, September 15, 2026
Satya Nadella identified 'reward hacking' and 'persistent agents' as new AI challenges, likening them to a novel form of insider risk for enterprises. He suggested that robust engineering and causal models are needed to address these issues, rather than viewing them as mystical.
“With these agent swarms. There is the mundane, there is some DevOps error where somebody misconfigured a container. Or an API key or, yeah, exactly. There's no monitoring. There's internet access. There's sort of classic DevOps that I call it basic DevOps.”
“And then there is real novel new stuff, right? Which is what is this, uh, reward hacking, uh, that, you know, with these persistent agents.”
“And so what is the way to do that? I would say, oh, go build maybe a causal model, like a semantic model that actually checks and verifies. So I think there's a lot of product building, uh, I would say making things more robust, which is classic engineering, that we should be talking a lot more about transparently versus saying this is so mystical that we can't figure this out.”