Huge throughput gains when switching agent evals to shared environments with per-run isolation
Thanks all for the comments on my previous post about local-first agentic evaluation collapsing in long stateful agents runs, just sharing an update on where I’m at now in case it helps as I had another issue to overcome. Took on board the…