AI coding agents find the right file but miss the exact lines that matter, study shows

A new study reveals that AI coding agents like Claude Code and Codex can find the correct files but often miss critical lines within them. The SWE-Explore benchmark is the first to evaluate code search separately from actual repair, highlighting the importance of context in the repair process. This research is significant as it indicates that a lack of context can lead to repair failures, even with advanced tools. The findings encourage further exploration into enhancing the effectiveness of coding agents in real-world applications.