AI Coding Boom Is Leaving Developers More Productive Than Ever, and More Exhausted

The rapid rise of artificial intelligence coding assistants is delivering dramatic productivity gains for software engineers, but many developers say those gains are coming with an unexpected cost: mental exhaustion.

From startup founders to engineers at leading AI companies, programmers increasingly describe a workday defined by faster output, relentless decision-making, and the pressure to keep pace with constantly improving AI tools.

The discussion gained momentum after Midjourney founder David Holz shared a candid observation on X about what he was hearing from fellow programmers.

"My programmer friends are all feeling extremely productive and also extremely drained with the latest coding models," Holz wrote.

He added that the trend made him feel "like something is wrong, and also that there might be a big opportunity," before asking other developers whether they had found strategies to make working with AI less mentally taxing on a day-to-day basis.

His comments quickly resonated across the software community, prompting engineers, AI researchers and technology executives to share both practical coping mechanisms and broader concerns about how AI is reshaping software development.

One recurring theme was that AI-assisted programming changes the nature of cognitive work.

Former Meta engineer Shuming Hu said that "vibe coding" - an emerging term used to describe rapidly directing AI models to generate code through conversational prompts rather than manually writing software - prevents developers from entering the deep state of concentration traditionally associated with programming.

Instead of becoming immersed in solving a single technical problem, programmers often find themselves continuously reviewing AI-generated code, refining prompts, correcting mistakes, and deciding among multiple implementation options.

That constant context switching, many developers suggest, may produce more software while simultaneously increasing mental fatigue.

Catherine Wu, Anthropic's Head of Product for Claude Code, said she deliberately limits her workflow at times to regain focus.

"I like focused work on a hard task with a single agent," Wu wrote.

Although she often runs dozens of AI agents simultaneously, she said concentrating on one difficult problem with a single AI assistant allows her to "get into the zone" and complete work more effectively.

Her comments point to an emerging challenge in AI-assisted development.

Modern coding tools can launch multiple autonomous agents that write code, debug applications, generate documentation, and perform software testing simultaneously. While those capabilities dramatically increase throughput, they also require developers to supervise several streams of work at once, shifting their role from writing code to managing AI-generated output.

Some developers believe the psychological effects may become even more pronounced as AI capabilities improve.

Former X and Cash App designer Brandon Kainoa Jacoby said the problem is "probably going to get worse before it gets better."

Rather than recommending more sophisticated AI workflows, Jacoby suggested stepping away from AI entirely for periods of time.

"I've noticed doing some sort of deep cognitive task, entirely away from any model, helps a tad," he wrote.

Other developers proposed similarly simple approaches, including spending time outdoors, looking at trees, taking walks, or playing with their children as ways to reset mentally after prolonged interaction with AI systems.

The conversation comes from a broader phenomenon increasingly referred to within the technology industry as "AI fatigue."

While artificial intelligence promises to eliminate repetitive programming tasks, many engineers say it has also introduced new forms of cognitive pressure. Instead of manually producing every line of code, developers now spend much of their time evaluating AI-generated suggestions, verifying correctness, comparing alternative solutions, and deciding when to trust automated output.

The result is less typing but often more continuous decision-making.

Concerns about AI fatigue have been building for months. In February, programmer Siddhant Khare published an essay arguing that AI fatigue is real but remains largely overlooked within the technology industry. The piece gained widespread attention among developers who said they recognized the same symptoms in their own work.

Some software engineers had earlier noted that the rapid pace of AI development had become overwhelming. New models, coding assistants, and software tools are released so frequently that many programmers struggle to keep up, creating anxiety that they may fall behind colleagues who adopt the latest technologies more quickly.

For some, the speed of change has produced a sense of workplace paralysis, where the constant arrival of new AI capabilities makes it difficult to settle on stable workflows or long-term development practices.

The pressure is also changing work habits.

Many developers say AI has significantly increased expectations around productivity, encouraging longer working hours as programmers attempt to maximize the advantages offered by increasingly capable coding assistants.

Ben South, a serial entrepreneur and former Vice President at Postmates, captured that mindset in his response to Holz's post.

"Even an hour of rest feels like a ton of productivity lost," South wrote.

As coding assistants reduce the time required to complete software projects, they also raise expectations about how much work can be accomplished in a single day. But that dynamic is said to carry risks of creating an environment where efficiency gains translate not into shorter working hours but into greater pressure to produce even more.

For technology companies, the discussion raises questions that extend beyond software engineering. The long-term success of AI in the workplace may depend not only on how much it improves productivity but also on whether workers can use powerful tools without experiencing sustained mental fatigue or burnout.

However, the conversation sparked by Holz suggests that the industry's next challenge may not be building more capable AI models. It may be designing workflows that allow humans to benefit from those systems without sacrificing the focus, creativity, and mental energy that have long defined effective software development.