SWE-bench co-author on building influential AI benchmarks: start with people
OfirPress · x · 2026-11-17
Kaggle launches its Evaluated interview series with SWE-bench co-author OfirPress, who shares how the benchmark went from a lukewarm launch to one of AI coding's most influential evals. His advice: start with people — find teammates to bounce ideas off — then tackle problems that seem impossible, and build tough benchmarks that matter for humans.
More from coding & agent
- YC launch CoverAgents turns broker emails into carrier quotes in 3 minutes, 90+ agencies live — ycombinator · 2026-11-17
- Open-source REQL builds code graphs to cut coding agents' token waste — cl0wnfire · 2026-11-17
- Using Figma MCP to let Claude create designs and prototypes straight into Figma — Tegadesigns · 2026-11-17
- Arkenos v0.6.12 ships set-and-forget autonomous research agents with free tier — _AustinCalvert_ · 2026-11-17
- Cloudflare open-sources Kitesurf browser engine for agents, ships DO and spend-limit updates — ritakozlov · 2026-11-17
- Zoho Desk launches Blended Agents: free to build, up to 30M free tokens monthly — shashib · 2026-11-17