Improve
The loops that make an agent better than it was last month — measure it, read its own traffic back, and turn finished conversations into something the next conversation can use. Every one of them ends at a human approval.
- Evals — the cases that grade an agent, and the gate that stops a regression.
- Refinement — production traffic read back as a ranked report of what broke.
- Outcomes and follow-ups — what the traffic was worth, and coming back on a promise.
- Knowledge — durable concepts extracted from finished conversations.
- Facts — distilled customer memory, approved one fact at a time.
- Harvest — skills mined from real traffic, promoted only if two gates hold.