AI coding tools are starting to look less like a productivity win and more like a compulsion, according to fresh survey data.
Sleep is often the first casualty. Nearly a third of respondents, 32%, have put off rest to keep coding, and 43% admit to working on after they meant to stop. Roughly 39% say the tools make it harder to switch off from work entirely.
Rootly co-founder and CTO Quentin Rousseau described the pattern in a LinkedIn post about watching Claude Code refactor a module at 2:47 a.m., calling agentic coding addictive and saying he needed medical help after he could not sleep. The dopamine hit of a correct agent run and the adrenaline of a failure keep the loop going, he argued.
The pull shows up even in career terms. Some 74% of respondents linked heavy AI use to a better shot at a raise or promotion, yet 51% said it made burnout more likely.
There is also the verification problem. The 2025 Stack Overflow Developer Survey put trust in AI answer accuracy at 29%, down from 40%, with 45% of developers frustrated by responses that were almost right but not quite. Output arrives fast, but confirming it is correct, secure, and maintainable still takes time.
When employers treat AI as a capacity multiplier, the time saved on individual tasks can vanish under larger pull requests and more generated changes, making AI-assisted programming as much a work-life balance issue as a tooling one.
