Community Poll

After this week, do you still trust AI coding benchmarks?

An independent benchmark from DeepSWE caught a frontier model recovering answers from git history, and flagged double-digit error rates on a popular leaderboard. So where does that leave the scores you have been using to pick a model?

Tap an answer to vote instantly. Results appear right here.