A field guide to writing your own eval harness
Why "vibes-based" testing collapses past 50 prompts, and the smallest harness that scales without becoming a second product.
APR 28, 2026 ·
#evals
SWE-bench Verified — leaderboard notes
"Worth reading the verified subset methodology before quoting any number from the headline board."
via MarkTechPost APR 20, 2026 ·