AI coding productivity is modest and uneven — 20-30%, not 10x¶
Insight: Stack Overflow's 2025 survey (49,000+ devs) shows 84% use AI tools but only 60% view them favorably — down from 70% in 2023. The biggest frustration, cited by 66%, is AI solutions that are "almost right but not quite," leading to time-consuming debugging. Studies show 20-30% productivity improvements, far from "10x" claims. 77% of developers do no whole-app "vibe coding" at work.
Detail: Addy Osmani synthesizes multiple studies: AI excels at greenfield projects but struggles with complex legacy codebases. Most developers use basic autocomplete, not full autonomous agents. Individual productivity gains aren't translating to team productivity. Trys Mudford adds a social dimension: within tech companies, it's taboo to voice healthy skepticism about AI productivity. When every engineer is incentivized to confirm 10x gains to their boss, CEOs relay false consensus — creating a feedback loop that distorts reality. The pressure to conform ("your livelihood is on the line") prevents honest discussion of actual results.
Sources
Related: existing entry "AI coding quality is a skill issue" in external/claude-code.md — COMPLEMENTS