LIVE · 100 answersLast run: 38h agoJudge: gpt-4o-mini
Run eval

A/B Compare

Compares the OLD prompt (production answers from session_logs) against the NEW prompt (regenerated with a candidate prompt). If the new prompt wins, click Export Markdown to download a self-contained report for the prompt-review agent.

No regeneration runs yet

Go to Prompt Lab, pick a weak session, and click Run with new prompt to trigger a regeneration via GitHub Actions. Once it completes, the comparison will appear here automatically.

Go to Prompt Lab →