The day’s strongest work treats AI coding as a governed engineering process. AutoMat tests scientific reproducibility, SAGA measures full agent latency, and RECAP records real prompt-to-edit traces.
AI coding agents now have enough local evidence to justify three concrete changes: recording prompt-to-edit traces during real development, testing scientific agents with claim-level reproduction tasks, and measuring…
The day’s strongest software-engineering work treats LLM coding as a controlled engineering problem: build harder artifacts, keep claims tied to evidence, and preserve human review.
LLM coding work needs tighter gates around urgent repairs, class-sized evaluation tasks, and public claims in research repositories.
The strongest work on this day makes software agents easier to constrain, inspect, and score. CodeStruct and SWE-Shield tighten code-agent evaluation around exact edits and design rules.
Concrete changes are showing up in three places: repository repair agents can work on named code entities and be judged on small valid diffs, patch evaluation needs design-constraint checks beyond test pass rate, and…