The Productivity–Burnout Paradox of AI Coding Assistants
AI coding assistants are software tools that use artificial intelligence models to generate, review, and modify code, promising faster development, fewer repetitive tasks, and streamlined workflows for programmers, but they are also reshaping how cognitive effort, attention, and emotional energy are spent during everyday software work. The paradox is blunt: engineers are shipping more, yet finishing their days exhausted. The rapid rise of these assistants is delivering dramatic productivity gains for software engineers, but many developers say those gains are coming with an unexpected cost: mental exhaustion. Instead of feeling empowered by extra capacity, many report feeling pressured to keep pace with accelerating tools and expectations. The industry narrative still celebrates productivity gains, but the lived experience is starting to sound less like a triumph of efficiency and more like a quiet wellness crisis in slow motion.
David Holz’s Warning Shot: ‘Extremely Productive and Also Extremely Drained’
The current debate was sparked when Midjourney founder David Holz publicly described what many programmers had been saying in private. He wrote that his programmer friends are “all feeling extremely productive and also extremely drained with the latest coding models”. In the same post, he admitted the trend made him feel “like something is wrong, and also that there might be a big opportunity” and asked whether anyone had strategies to make day‑to‑day AI coding feel better. That combination of concern and opportunism captures the moment: leaders see a warning sign about developer exhaustion, and at the same time, a business opening for tools or practices that reduce AI tools fatigue. The replies he received were telling. Developers did not call for more automation; they called for relief—flow states, fewer agents, and breaks away from screens. That is not the language of a satisfied workforce.
From Flow to ‘Vibe Coding’: Why Productivity Gains Feel So Draining
Industry observers are clear that the exhaustion is not imaginary; it is baked into how AI coding assistants change the work. Former Meta engineer Shuming Hu argued that “vibe coding” prevents developers from entering the deep flow state traditionally associated with programming. Instead of immersing in a single problem, developers now spend much of their time evaluating AI‑generated suggestions, verifying correctness, comparing solutions, and deciding when to trust automated output. Modern tools can launch multiple autonomous agents that write code, debug, generate documentation, and run tests at the same time, dramatically increasing throughput but turning the developer into a full‑time supervisor of parallel workstreams. That means less typing and more continuous decision‑making. Cognitive science is clear: constant context switching is tiring. AI tools fatigue is not about staring at a model; it is about being forced into a frantic, multi‑threaded mental mode that software work did not require at this scale before.
Escalating Expectations and the New Shape of Developer Burnout
Productivity gains rarely stay neutral. Once AI coding assistants prove they can accelerate output, expectations rise to match. Many developers say AI has significantly increased expectations around productivity, encouraging longer working hours as they try to maximise the advantages of increasingly capable tools. One entrepreneur captured the mindset bluntly: “Even an hour of rest feels like a ton of productivity lost”. At the same time, new models and assistants are released so frequently that programmers struggle to keep up, creating anxiety about falling behind and a sense of workplace paralysis. Instead of stable workflows, developers face moving targets and endless onboarding to new AI tools. This is classic developer burnout in a modern wrapping: not only more work, but more pressure, more change, and more fear of being obsolete. The industry is discovering that efficiency can be weaponised against its own workforce when there are no guardrails on pace or expectations.
Designing AI Workflows That Protect Human Focus and Energy
The emerging conversation hints at what a healthier relationship with AI coding assistants might look like. Catherine Wu, who works on a leading coding assistant, said she prefers “focused work on a hard task with a single agent,” even though she often runs dozens of agents. Developers responding to Holz suggested stepping away from AI for stretches, doing deep cognitive tasks without any model, or simple offline resets like looking at trees or playing with their children. These are small acts of resistance against a culture that equates constant acceleration with success. Concerns about AI fatigue have been building, with one programmer arguing in February that the phenomenon is real but overlooked. The next frontier for AI in software development is not only more capable models; it is workflow design that respects human limits. As one discussion put it, the long‑term success of AI at work may depend on whether humans can use powerful tools without sliding into sustained mental fatigue or burnout.
