Ask HN: For those of you building AI agents, how have you made them faster?

2 pointsposted 7 hours ago
by arkmm

Item id: 46637298

1 Comments

codingdave

4 hours ago

To start with, I'm not coordinating across multiple systems or chaining LLM calls. I write all data to a central data store, run a state machine on it, and each LLM call then operates independently.

At that point, you can measure and optimize each call as needed and it is fairly straightforward.