Four in five developers say their use of AI coding tools has felt more like dependence than an advantage, according to a survey published by Coddy Tech, a programming training company. The survey of 305 developers reveals growing concerns that while these tools remain useful, their habitual use is undermining programmers' problem-solving abilities, expanding workloads, and creating unhealthy work habits. What began as productivity aids like GitHub Copilot and Claude Code are now raising questions about whether they're enabling burnout rather than preventing it.

The behavioral patterns described in the survey show how deeply AI tools are reshaping work boundaries. More than two-fifths of developers—43%—continue coding with AI after hours even when they intended to stop, while 32% have delayed sleep to keep working. Another 39% reported that AI tools have made it harder to disconnect from work entirely. Meanwhile, heavy AI users face competing pressures: 74% said intensive use made them more likely to earn a raise or promotion, but 51% also said it increased their likelihood of burning out. Trust in AI accuracy has also declined sharply, dropping from 40% in previous years to just 29% this year, according to the 2025 Stack Overflow Developer Survey cited in the report. That same survey found 45% of respondents frustrated by AI answers that were "almost right, but not quite," creating convincing output that leads to difficult debugging work. Overall positive favorability toward AI among developers fell from 72% to 60% year over year, even as adoption climbed to 80% of developers now using these tools in their workflows.

The report describes what Quentin Rousseau, CTO and co-founder of AI-powered incident report company Rootly, calls the addictive nature of agentic coding. "When the agent gets things right, you get a dopamine hit. When it fails, you get an adrenaline rush," Rousseau wrote in a LinkedIn post quoted in the report. He described being unable to sleep at 2:47 a.m. while watching an AI agent refactor code—not for a deadline or outage, but because he couldn't stop—and eventually sought medical help. The report frames this as a new kind of burnout where "watching an agent's work is passive enough to feel like rest, active enough to keep you hooked." Developers must still understand requirements, spot when generated code conflicts with system architecture, test edge cases, address security risks, and own production consequences—creating what the report calls "verification debt," where output arrives quickly but programmers remain stuck confirming whether it's correct, secure, and maintainable.

The tension arises because organizations often treat AI as a way to multiply developer capacity, the report explains, leading to pressure on workers to ship more features, close more tickets, and complete more reviews in the same hours. That pressure erases time saved on individual coding tasks and shifts the burden elsewhere: larger pull requests, more generated changes to inspect, more dependencies to validate, and more operational risk to manage. The result turns AI-assisted programming into as much a work-life balance issue as a tooling issue. Teams using agents to remove routine work may see genuine benefits, the report states, while those accelerating every part of the software production pipeline risk creating "a faster, more relentless version of the same job." The question going forward, the report concludes, is no longer whether AI can write code—it can—but whether developers can still decide when the workday and the agent loop end. Organizations that push for maximum velocity without addressing verification overhead and always-on work patterns may find their productivity gains come at the cost of sustainable careers. The challenge for engineering leadership will be drawing lines that preserve the advantages of automation without letting those same systems dictate when human judgment can rest.