THE CLAIM. AI coding tools deliver massive productivity gains, justifying industry-wide adoption at 93 percent and billions in corporate AI spending.

THE CHECK. six independent research efforts converge on roughly 10 percent organizational productivity gains at the system level, against vendor claims of 50 percent or more. METR's randomized controlled trial with 16 experienced open-source developers across 246 tasks found that developers using AI took 19 percent longer, while believing they were 20 percent faster, a 39-point perception gap between subjective experience and measured reality. 56 percent of AI suggestions were rejected. Faros AI data from 10,000 developers shows 98 percent more pull requests merged alongside a 91 percent increase in review time and a 9 percent rise in bugs. Uber blew through its entire 2026 AI coding budget in four months after ranking engineers on an internal leaderboard by tool usage, and COO Andrew Macdonald said it was very hard to draw a line to actually producing 25 percent more useful consumer features.

THE PATTERN. adoption driven by perception, funded by assumption, measured by the vendor. When an independent stopwatch shows up, the gains shrink or reverse.