VQV Signal
According to a post on X by @trevin, which summarizes USC research, Claude Code and Codex were repeatedly asked to "refactor further" on code that was already clean. The post says the tools changed the code 85% of the time, rewriting it 8x to 31x the original and mostly adding code. The post was reshared by @simp4eloquence, and the provided snippet does not include the study's full methodology.
According to a post on X by @trevin, which summarizes USC research, Claude Code and Codex were repeatedly asked to "refactor further" on code that was already clean. The post says the tools changed the code 85% of the time, rewriting it 8x to 31x the original and mostly adding code. The post was reshared by @simp4eloquence, and the provided snippet does not include the study's full methodology.
AI-assisted summary based on the listed source.
@trevin summarizes USC research in which Claude Code and Codex were repeatedly asked to "refactor further" on code that was already clean: 85% of the time they changed it anyway, rewriting 8x to 31x the original and mostly adding code. @simp4eloquence reshared it this morning. Original post on X Section: Trends &...
Developers and teams using AI coding agents may want to give them a specific goal instead of an open-ended request to keep refactoring. The research summarized in the post suggests that without one, agents may churn already clean code and add volume, which could make reviews harder. Readers should look for the full USC study to check its methods and how widely the results apply.
VQV organizes public signals from inspectable sources. It does not independently verify the underlying report.
Signal Strength reflects source quality, relevance, freshness and evidence. Public Interest helps organize discovery; it is not proof of truth.
VQV surfaced this signal because it is recent, relevant to AI Coding Tools, connected to Scobleizer News Pulse (X).
No login, cookies, social SDKs, or automatic posting.
