The people building the AI boom are running on empty. According to Stack Overflow's 2026 Developer Survey, released Tuesday, nearly half of 30,903 developers surveyed in 169 countries reported being complacent in their current jobs, while a third were unhappy, and just over 22% were happy. Among the complacent group, burnout or tech fatigue was the leading reason, cited by 21% of respondents. The findings come amid an environment where software engineers' jobs are changing at breakneck speed. AI coding has taken over a greater share of their daily work — reducing some developers to button-pushers and human meat proxies, contributing to an identity crisis among some engineers, and fueling "off-the-charts" anxiety among tech workers.

The malaise comes alongside a major shift in how software gets made. Stack Overflow found that almost two-thirds of respondents use AI coding assistants or agents, while 62% use general-purpose AI chat tools. Among daily AI users, 31% said they use them for four hours or more.

The findings echo what Annie Wright, a Bay Area psychotherapist and executive coach whose practice is dominated by tech workers, told Business Insider in August. Her clients had moved beyond ordinary burnout, she said, to questioning their identity and future as automation reshapes the work they have built their careers around. "Three years ago, people would walk in saying, 'I'm exhausted, burned out beyond belief,'" Wright said at the time. "Now they walk in saying, 'I'm exhausted. I'm burned out beyond belief.

I don't know who I am if my work gets automated.'" Marc Backes, a senior software engineer at Directus, said AI has made his work less fulfilling. He used to enjoy working through a feature or tracing a technical problem to its source; now, he spends far more time asking AI agents to do it. "It's definitely less fulfilling," Backes told Business Insider.

"I always enjoyed drafting out the best ways to implement a new feature and really think about it." The pressure is not coming directly from his manager, he said, but from an industry moving faster and treating writing — or even reading — code by hand as "the old way." "If I don't 'keep up,' I will be 'left behind,'" Backes said. "And keeping up with AI is virtually impossible." Not everyone feels diminished by the shift, though.

Simon Willison, founder of the Datasette open-source project, said AI coding tools have made his work "a lot more fulfilling" by allowing him to take on projects that once might have required a small team. Willison said that being his own manager shields him from many of the pressures facing employed developers, though he still finds the pace of AI releases exhausting. "As an entrepreneur and product developer, I know that any space I operate in will be significantly more competitive now because the cost of developing software has dropped so much," he said.

Still, Willison said he was optimistic about his career. "Using coding agents well is a deep skill," he said, one that requires both software-engineering knowledge and experience with the agents, "and I have both." Despite many using AI coding tools extensively, developers don't necessarily trust their output. The survey found that under 10% of respondents said they were prepared to trust AI for important decisions, while almost 80% said source attribution was important or very important when evaluating AI-generated code or technical answers.

Prashanth Chandrasekar, Stack Overflow's CEO, said developers' skepticism was becoming a "critical safeguard" as agents took on more work. "Trust is not a given when it comes to developers," he said. "It is earned through transparency, source attribution, and rich technical context."