{"last_refreshed":"2026-10-30","cadence":"When a new lab record is committed. No separate clock in the repo.","source":"Marketing lab challenger record, tested 2026-10-30.","winners":[{"id":"design-bakeoff","date":"2026-10-30","name":"pbakaus/impeccable","metric":"Score out of 10 on one landing-page draft","result":"8.1","verdict":"winner"},{"id":"dspy-headlines","date":"2026-10-30","name":"DSPy","metric":"Headline and meta score on 20 held-out Notebook pages","result":"49.67 to 65.08","verdict":"winner"},{"id":"mem0-recall","date":"2026-10-30","name":"Mem0","metric":"Lesson recall at 3 versus a keyword lookup","result":"0.59 versus 0.36","verdict":"winner"}],"losers":[{"date":"2026-10-30","name":"Nutlope/hallmark","metric":"Score out of 10 on one landing-page draft","result":"7.9","verdict":"loser"},{"date":"2026-10-30","name":"elayadesign/ai-design-skills","metric":"Score out of 10 on one landing-page draft","result":"7.3","verdict":"loser"},{"date":"2026-10-30","name":"emilkowalski/skills","metric":"Score out of 10 on one landing-page draft","result":"7.0","verdict":"loser"},{"date":"2026-10-30","name":"GEPA","metric":"Prompt score on a local 1.5B model","result":"0.1283 to 0.1284","verdict":"loser"},{"date":"2026-10-30","name":"OpenEvolve","metric":"SPY seed Sharpe over 12 iterations","result":"0.789 not beaten","verdict":"loser"},{"date":"2026-10-30","name":"Agent Lightning","metric":"Topic-agent accuracy","result":"0.55 to 0.475","verdict":"loser"}],"note":"These are the tested rows in the lab record. Queued items with no result are not listed as winners or losers."}