Programming and fatigue
Nate Meyvis argues that programmers feeling more tired since adopting AI tools may be doing higher-judgment work rather than burning out. He admits he can't prove it, but thinks the fatigue is more likely a sign of productive effort than a cue to change how they work.
- The fatigue is plausibly healthy: AI handles the repetitive and obvious parts, leaving the human with harder judgment calls, which are more tiring.
- He compares AI-assisted programming to juggling many windows in online poker, and says the fatigue feels similar.
- He admits the view is optimistic and unproven, and that AI doesn't clearly replace only the less tiring work, since log reading and research can be taxing.
Nate Meyvis, writing on his blog, says programming now tires him more than it used to, and that many other programmers report the same thing. Where most people read that fatigue as a warning sign that their workflow is broken, he reads it as a good sign. His reasoning: if he's mentally drained, it probably means he's been doing a lot of real work and spending his time on tasks that need judgment. AI handles the repetitive and obvious parts, which leaves him with the harder, less routine decisions, and those are simply more tiring. He thinks this fits an optimistic picture of human-AI teamwork in which the person's role shifts toward higher-judgment work rather than disappearing.
He is careful about the limits of this view. He doesn't want to romanticize burnout, and he admits he can't prove the fatigue comes from healthy effort rather than something he should fix, such as frequent context-switching. He also points out that some tasks AI is good at, like reading logs and doing certain research, are themselves taxing, while some tasks left to him are easy and routine, so the idea that AI only takes the less tiring work isn't clearly true. His main comparison is online poker, where he juggled 8 to 12 windows at once. Running many active threads in a programming session feels similar, and the resulting tiredness feels similar too. He calls his own view optimistic and conjectural, and acknowledges that others might find it delusional.
On workflow, he describes himself as somewhat old-fashioned. He automates what he reasonably can, but he follows Gall's Law (a complex system that works evolved from a simple one that worked) and doesn't want to jump into a full "Software Factory" model of AI-driven development. His concern is that such setups make it hard to stay flexible about when to override or guide the AI, and he isn't convinced they'll improve his work. He says he's eager to experiment with heavy AI use but finds it hard to predict where human intervention helps most. His closing advice is to look after yourself, but not to assume that more fatigue means you're doing something wrong.