Guide, Verify, Solve — Anirban Chatterjee, Sonar
CMU study found AI coding tools yield only a 3-month productivity spike before quality debt erases gains
“there was, in fact, a temporary spike in productivity... but it lasted about 3 months and then it went back down”
A Carnegie Mellon study of GitHub projects found that AI-assisted coding (using Cursor as a proxy) produced only a temporary ~3-month productivity boost, after which velocity returned to baseline while static analysis warnings and code complexity persistently increased. Sonar is citing this research to argue that AI coding workflows require embedded quality guardrails to sustain gains. The finding directly challenges the prevailing narrative that AI coding assistants deliver durable productivity improvements.