I have been using claude/gemini/gpt heavily for 6-8h a day for research and error catching. Incredibly useful to find logical errors I miss, challenges my assumptions, speeds up thinking like a lot. It even improves my self awareness.

But I am hitting my biological limits. After 6+ hours, I'm mentally exhausted. Eyes and back hurt, brain fog, can not engage with real people. The AI doesn't get tired - I do. I know that gym is one part of the answer. But also:

  I am building a tracking system for this                                                                                                       
  - 4h daily limit (hard stop, biological threshold)                                                                                              
  - challenge our convergence (track if AI just agrees vs actually challenges)                                                                      
  - energy economy metric (does it actually preserve or drain capacity?)                                                                              
                                                                                                                                                  
  AI has crazy capabilities and it can be genuinely helpful for some cognitive work. But it feels like a trap if I don't watch my own boundaries.                                                               
                                                                                                                                                  

Questions for you guys: - What's your sustainable AI usage? (hours/day that doesn't wreck you) - Do you track some aspects of your collaboration? (metrics, feelings, hard limits?) - How do you know when to stop? (before exhaustion or after?) - Am I overthinking this? (should I just... use less AI? Although it makes me right now 100 times more productive?)

Building this boundary system partly because I can't trust myself to stop without forcing functions and bias the system further without external feedback. Anyone else need this or just me?

Brutal feedback wanted. If my approach is wrong, tell me. Or just tell me how you are doing it. Thank you in advance!