It's the current industry buzzword for action-oriented LLMs.

Instead of generating text, tightening/formalizing the loop around planning, executing, analyzing results, and replanning.

As far as buzzwords go, it's far from the worst, as it captures the essentials -- creating semi-autonomous agents.

Based on the context Ilya is not referring to that kind of agent. He’s referring to something more fundamental (which I was curious about, too).

Example of 'agentic': https://blog.google/technology/google-deepmind/google-gemini...

When it autonomously performs tasks on behalf of the user, without their intervention. It sets goals, plans actions, etc. by itself.