[chatgpt-roadmap] ## Future capability Track worker performance so the fleet can learn which models are most useful for different task types. Potential signals: - Accuracy or evaluation score - Latency - Failure/timeout rate - Agreement with other workers - Arbiter/verifier outcomes - Cost where applicable **Not required for PR #2.** This is a post-MVP capability to revisit after the current MCP fleet implementation is stable and tested.
[chatgpt-roadmap]
Future capability
Track worker performance so the fleet can learn which models are most useful for different task types.
Potential signals:
Not required for PR #2. This is a post-MVP capability to revisit after the current MCP fleet implementation is stable and tested.