80% Of Developers Call AI Coding Addictive. They Are Not Wrong. The Dopamine Is The Point.
A Coddy Developer Survey found that 80 percent of developers find AI coding more addictive than helpful, with a new form of burnout emerging. Quentin Rousseau, CTO and co-founder of Rootly, described the compulsion in a LinkedIn post: at 2:47 a.m., with no outage to debug and no deadline, he found himself watching Claude Code refactor a module simply because he could not stop. The tools in question include GitHub Copilot, Claude Code, and Cursor.
This is a textbook case of variable reward reinforcement, the same behavioral mechanism that drives slot machines and social media scrolling. AI coding tools generate fast, satisfying output that triggers dopamine release. The mental model here is compulsion loops in productivity software. The tool removes friction so effectively that the user cannot disengage, even when the task has no productive purpose. Recognizing this pattern is the first step toward managing it. The technology is not the problem. The reward schedule is.
Coddy conducted the developer survey. Quentin Rousseau, CTO and co-founder of Rootly, provided the firsthand account on LinkedIn. The tools named are GitHub Copilot, Claude Code, and Cursor, all widely used across professional development teams.
- Open a free AI chatbot and ask it to perform a small creative task, such as rewriting a paragraph from your resume in three different tones.
- Notice the impulse to immediately ask for another variation, then another, without stopping to evaluate whether the output is actually useful.
- Set a timer for five minutes. When it goes off, close the tab. The goal is to observe your own compulsion loop in real time. If you feel resistance to closing the tab, you have just experienced what 80 percent of developers reported.