Agents differ from models via the machinery: tools, memory, loops

vishalmisra · x · 2026-09-27

In a follow-up to his MIT talk, Vishal Misra argues that agents are different mostly because of the machinery around the model: tools, memory, permissions, feedback, and repeated calls.

The model itself is still doing next-token prediction; it's the surrounding system that turns those predictions into actions, state, and loops.

Related event: MIT Talk Explains LLMs from First Principles, No Transformer Needed(2 posts)→

Original post →

More from coding & agent

coding & agent channel →