The New Productivity Paradox in Software Engineering
AI coding assistants are tools that automatically generate, debug, and document software code, dramatically increasing software engineer productivity while simultaneously changing developers’ day-to-day cognitive workload in ways that can cause surprising mental fatigue and burnout. The headline story of the AI boom is speed: engineers using tools like Claude Code are seeing their productivity increase by two to three times, according to Anthropic’s head of growth Amol Avasare. Yet many of those same engineers describe feeling “extremely productive and also extremely drained” when working with the latest coding models, as Midjourney founder David Holz bluntly put it. That tension is not a temporary growing pain; it is the core paradox of AI-assisted programming. We are getting more code and fewer keystrokes, but also more pressure, more context switching, and more decision fatigue. The industry has mistaken more output for better work—and developers’ brains are paying the price.
From Crafting Code to Supervising Machines
AI coding assistants promise to eliminate repetitive programming tasks, but they have replaced typing with nonstop judgment calls. Instead of entering a deep flow state to solve one problem, developers now juggle prompts, review AI-generated code, correct mistakes, and pick between multiple implementations. Modern tools can spin up several autonomous agents that write features, run tests, and generate documentation at once, forcing engineers into a kind of human air-traffic control for software. Former Meta engineer Shuming Hu points out that “vibe coding” — steering code through conversational prompts — often prevents the focused immersion that used to define quality programming time. The supposed gain is throughput; the hidden cost is constant context switching and mental fatigue. As coding assistants cut project times, managers rarely respond by shortening workdays. Instead, expectations rise: if a day of work can produce twice the code, then pausing for rest starts to feel like lost productivity.
Why People Skills Just Became Core Engineering Skills
As AI writes more code, companies are asking software engineers to review, direct, and manage AI-generated work rather than produce every line themselves. That shift quietly rewrites the job description. Engineers are being asked to take on more product management tasks, coordinate stakeholders, and own cross-functional outcomes as “mini PMs” on smaller projects. The rise of this hybrid “product engineer” role means technical brilliance without people skills is starting to look like a liability. Kent Beck, a legendary software figure, needles the profession with a harsh truth: programmers are often praised for technical depth while excused for lacking empathy, emotional regulation, and tact, which he calls some of their more “hideous” qualities. Now the joke is on them. AI has automated much of the individual contributor coding grind, so the scarce skill is no longer algorithmic cleverness but product judgment, communication, and the ability to empathize with teammates and users. The engineers who refuse this evolution will find themselves supervising machines without influencing the direction of the work.
Burnout, AI Fatigue, and the Illusion of Job Security
Developers hoped AI coding assistants would mean less toil and more sanity. In practice, they are getting “AI fatigue” instead. Programmers now spend much of their time evaluating AI suggestions, verifying correctness, comparing alternatives, and deciding when to trust automated output, which leads to continuous decision-making and mental exhaustion. Concerns about this pattern have built for months; in February, programmer Siddhant Khare argued that AI fatigue is real but largely overlooked, resonating with peers who saw their own symptoms in his description. At the same time, new models appear so quickly that many engineers feel pressure to keep up with every release, fearing they will fall behind coworkers who adopt the latest workflows sooner. Instead of improved work-life balance, faster code has created a culture where “even an hour of rest feels like a ton of productivity lost,” as entrepreneur Ben South observed. Job security is no longer about writing more code; it is about sustaining mental health in a system that treats human attention as an infinite resource.
What Developers Must Learn Next to Survive the AI Era
Some developers warn that the psychological effects of AI-assisted coding are “probably going to get worse before they get better,” suggesting the industry has not yet understood its new cognitive risks. The discussion sparked by David Holz’s post hints that the next challenge is not building more powerful models but designing workflows that let humans benefit from them without losing focus, creativity, or mental energy. That requires developers to grow beyond pure technical skills. Engineers are being asked to act as product thinkers, communicators, and coordinators — roles where people skills are not optional but central to impact. Kent Beck describes this as a “cosmic practical joke”: programmers were told that mastering computers was enough, only to discover that their real influence is capped by their ability to communicate and empathize. In the coming years, the most resilient developers will be those who treat AI coding assistants as powerful but demanding collaborators, set boundaries to avoid burnout, and invest in people skills for developers with the same seriousness they once reserved for learning frameworks and languages.
