I burned out too. Late last year I was running 4-5 projects in parallel with AI, always wanting every spare second utilized. Put up some amazing output numbers. Then hit a wall.
Since then I've pulled back to 2-3 at a time - thats sustainable for me. But I had to build systems to make it work: larger chunked tasks so I'm not constantly context switching on small stuff, adversarial LLMs where Claude writes and Codex judges so reviews are more solid, tooling to track whether I'm shipping real complexity or just noise. (Thats actually why I built https://gitvelocity.dev - wanted to measure my own output honestly, not just feel productive.)
Two things I've landed on:
First - the "don't context switch" adage needs updating. It was true when switching meant reloading everything in your head. With AI handling more of the implementation, the switching cost is lower. My wife juggles 10 things managing our kids schedules/appointments and shes way better at it than me - not wired differently, just practiced. I think we can build that muscle.
Second - were still measuring productivity the old way. Commits, PRs, lines. AI makes volume trivial so expectations ratchet up. If we measured complexity and value instead, the pressure to churn would ease.
Burnout is real. But some of this is growing pains, not permanent condition.
Mike, your issue is exactly what we want to solve. We thought about the current assumptions of how people order today:
1) Figure out what to eat
2) Determine if it's worth the money
On Munchery, we flip it entirely around:
1) A chef figures out what you eat based on the best ingredients of the season
2) The meal always costs the same
Or in foodie terms, Munchery is French prix fixe + Japanese omakase. We think that not only brings great value to customers, but it's just plain exciting.