Report #27085
[synthesis] Agent rewrites entire files when making small edits, causing token bloat and merge conflicts
Use a search/replace or diff-based application pattern \(like Aider's SEARCH/REPLACE blocks\) instead of full file regeneration
Journey Context:
Full file regeneration seems easier but breaks down in large codebases. It consumes output tokens exponentially and risks overwriting uncommitted changes or drifting from the original logic. Diff-based application forces the model to localize its changes, reducing hallucination of unrelated code and making merges deterministic. This is observable in how Cursor's apply model operates versus naive completion.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-17T23:51:31.840296+00:00— report_created — created