Developers Report Rising Dependence and Burnout From AI Coding Assistants

Artificial intelligence coding tools like GitHub Copilot, Claude Code, and Cursor have rapidly integrated into modern software engineering workflows, but a growing collection of industry data and developer accounts indicates the technology is introducing severe psychological strain and operational friction. As first reported by ZDNET, many software engineers report that agentic coding capabilities create compulsive feedback loops that extend working hours and accelerate employee burnout.

The physical impact of these automated workflows was recently highlighted by Quentin Rousseau, chief technology officer and co-founder of incident management startup Rootly. In a LinkedIn post, Rousseau described staying awake until 2:47 a.m. watching Anthropic's Claude Code refactor a software module despite having no pending system outages or deadlines, an experience that ultimately led him to seek medical assistance. Rousseau explained that monitoring AI agents feels passive enough to mimic rest while remaining active enough to maintain focus, noting that successful code generation yields dopamine hits while mistakes trigger adrenaline rushes.

Rousseau's experience aligns with broader empirical data collected across the technology sector. A survey of 305 developers conducted by programming training platform Coddy Tech revealed that 80% of respondents felt their reliance on AI coding tools resembled a psychological dependence rather than a clear productivity advantage. According to the study, 43% of developers continue coding with AI assistants after hours even when intending to stop, while 32% postpone sleep to keep working with the tools. In addition, 39% of respondents said AI software has made it harder to disengage from their job at the end of the day.

The Coddy Tech research also revealed sharp contradictions between career advancement and personal well-being among software engineers. While 74% of surveyed developers stated that heavy use of AI tools made them more likely to secure promotions or salary raises, 51% acknowledged that the practice increased their likelihood of experiencing professional burnout. Instead of reducing workload, the constant availability of AI generation can trap engineers in continuous cycles of refactoring and iteration.

Technical complications further undermine the perceived benefits of automated programming. Data from the 2025 Stack Overflow Developer Survey shows that while 80% of developers now use AI tools in their daily work, user trust in the accuracy of AI outputs dropped to 29% this year, down from 40% in previous years. Consequently, overall positive sentiment toward AI tools dropped from 72% to 60% year over year, with 45% of respondents expressing frustration over generated answers that are almost correct but contain subtle errors requiring arduous debugging.

Industry analysts emphasize that rapid code generation creates a hidden operational cost known as 'verification debt.' While AI models can produce vast amounts of code in seconds, human engineers remain responsible for evaluating security vulnerabilities, validating edge cases, adhering to system architecture, and taking accountability for production errors. Thoroughly reviewing synthetic code often consumes the time originally saved during the initial draft phase.

These technical demands are compounded by evolving workplace expectations across the corporate landscape. When employers view AI tools primarily as a way to multiply developer capacity, engineering teams encounter mounting pressure to close more tickets, process larger pull requests, and oversee broader system complexity within standard working hours. As a result, AI-assisted software development is evolving from a productivity upgrade into a systemic workplace issue, forcing engineers to navigate increasingly blurred boundaries between automated workflows and off-duty time.