80% of Developers Find Code AI Addictive

The adoption of AI tools for coding is progressing, but their appreciation is declining, and trust in their responses has fallen to 29% according to Stack Overflow 2025. According to Coddy Tech, a majority of developers perceive an addictive usage, with some struggling to disconnect and experiencing shortened nights. One CTO even described insomnia related to following a refactoring agent. Beyond the promised speed gains, a "verification debt" is settling in and reconfiguring the workload.

Stack Overflow 2025: 80% Adoption, Trust at 29%, Favor at 60%

According to Stack Overflow 2025, 80% of developers have integrated AI tools into their workflows. At the same time, trust in the accuracy of these systems has dropped from 40% in previous years to 29% this year, and programmer favorability towards AI has decreased from 72% to 60% year-over-year. 45% of respondents report frustration with "almost correct, but not quite" answers, a gray area that complicates corrections. Developers also need to ensure that the generated code aligns with the architecture, covers edge cases, does not introduce vulnerabilities, and remains maintainable. This "verification debt" shifts the burden from generation time to ensuring that the output is actually usable.

Coddy Tech: 80% Perceive Addiction, Extended Hours

According to Coddy Tech, four out of five developers feel their use of AI is closer to an addiction than an advantage. 43% say they continue coding with AI after work hours despite intending to stop, 32% have postponed their sleep to continue, and 39% find it more difficult to disconnect. Respondents consider these tools useful, while fearing that addiction may weaken their own problem-solving approach, increase their workload, and encourage an unhealthy relationship with work.

A CTO Shares a Sleepless Night with Claude Code

Quentin Rousseau, CTO and co-founder of Rootly, recounted on LinkedIn that he found himself at 2:47 AM watching a Claude Code agent refactor a module, with no urgency or failure to manage, explaining that he couldn’t bring himself to stop. He describes an addictive experience, marked by a dopamine rush when the agent succeeds and adrenaline in case of failure, to the point of consulting a doctor for sleep disorders. The use of AI in coding can transform software development into a continuous loop of implementations, rewrites, and optimizations. Some developers worry that stopping would mean leaving work unfinished.

Organizational Pressure and Shifted Burden with Agents

The time saved by AI can be offset by the corrections needed for the generated code. Such tools can also promote workaholism and burnout among developers. The addiction described by Coddy Tech could be exacerbated by how employers interpret the results produced by AI. If an organization views AI as a capacity multiplier, developers may feel pressured to deliver more features, close more tickets, and complete more reviews within the same time frame, which can shift the burden towards larger pull requests, more changes to inspect, more dependencies to validate, and more operational risks to manage. Benefits are mentioned when agents eliminate routine tasks. Conversely, trying to accelerate every step of software production with agents risks producing a faster and more relentless version of the same work.