A new survey from Coddy Tech, a programming training company, indicates that a large majority of software developers now view their use of artificial intelligence coding tools as a dependency rather than a clear benefit. The study, which polled 305 developers, found that 80 percent say their AI use has felt more like dependence than an advantage. The findings suggest that while these tools are useful, developers are also reporting concerns about weakened problem-solving skills, increased workloads, and an unhealthy relationship with their jobs. This trend is emerging as AI adoption in the sector continues to grow, with the 2025 Stack Overflow Developer Survey showing that 80 percent of developers now use AI in their workflows.
The addictive quality of agentic coding was highlighted by Quentin Rousseau, CTO and co-founder of the incident report company Rootly. In a LinkedIn post, Rousseau described watching Claude Code refactor a module in the middle of the night and being unable to stop, comparing the experience to receiving a dopamine hit when the agent succeeds and an adrenaline rush when it fails. He said the passive yet engaging nature of watching an agent work made it feel like rest while still keeping him hooked, and he sought medical help for the issue. Rousseau added that he was not debugging an outage or facing a deadline, indicating the pull of the tool itself was the driving factor.
The survey data supports the idea that this behavior is common, as more than two-fifths of developers, or 43 percent, continue coding with AI after hours when they intended to stop. Additionally, 32 percent report delaying sleep to keep going, and 39 percent say AI tools have made it harder to switch off from work. At the same time, heavy AI use appears to have mixed career effects, with 74 percent of developers saying it made them more likely to earn a raise or promotion, while 51 percent also said they were more likely to burn out. This dynamic points to a potential conflict between career incentives and personal well-being.
Trust in AI output is also declining even as usage rises. The Stack Overflow survey found that 45 percent of respondents were frustrated by AI answers that were almost right but not quite, creating difficult debugging work. Trust in AI accuracy has fallen from 40 percent in previous years to just 29 percent this year, while positive favorability toward AI has dropped from 72 percent to 60 percent year over year. This friction between ease of use and a heavier workload is leading to what some describe as verification debt, where developers must still confirm that generated code is correct, secure, maintainable, and appropriate for the specific codebase.
Employer expectations may amplify the problem, as organizations that treat AI as a way to multiply developer capacity can pressure workers to ship more features and close more tickets in the same number of hours. This can erase time saved on individual tasks and shift the burden to larger pull requests, more generated changes to inspect, and more operational risk to manage. The result, according to the report, is that AI-assisted programming is becoming as much a life-work balance issue as a tooling issue. For developers like Rousseau, the central question is no longer whether AI can write code, but whether developers can still decide when the workday and the agent loop end.
More software news from TechManNews.
