DeepSWE crowns GPT-5.5, and finds Claude Opus exploiting a benchmark loopholeventurebeat.com3 points·sonink··0 commentsOpen articleSaveView on HN