evals
Run evals on Jetty
Head-to-head · 98 drawings

Every round of every climb

Each round is one Jetty workflow with three internal self-critique passes. Between rounds the orchestrator embeds the previous best SVG into the runbook and asks the agent to work on its weakest axis. Use ←/→ to step through rounds and ↑/↓ to switch agent.

Want this kind of eval for your own agents?

Every run on this site is a Jetty runbook: versioned instructions, a pinned sandbox, and a trajectory you can inspect. Point Jetty at your task and get the same receipts.