Michael Hummel on Mythryl: “Small accuracy gains compounding into longer agent tasks assumes the…”

Small accuracy gains compounding into longer agent tasks assumes the only failure term is the model. Run an agent inside someone else's system and each step also survives an expired token, a changed API, a record some other process locked. The model got smarter - the environment didn't.

x.comEthan Mollick (@emollick) on XI think this will turn out to be wrong, and not just because I suspect there are greater returns to more intelligent models than people expect Economic value comes from agents, not chatbots. And accu...

14/8/26, 5:26 pm0 reposts0 likes

Post