Replying to @Jason_A_Scharf @johnternus
Probably there for F1 this weekend and the timing lined up. I don鈥檛 think it鈥檚 that deep
Replying to @catalinmpit
Share the prompt you used. Otherwise going to chalk this up as a skill issue
Replying to @charliermarsh
Reading for long periods of time has been the only thing that has helped me restore my attention span
Between agents, socials, and everything fighting for my dopamine, its a battle to keep it in tact
Have to put the phone somewhere else too because the brain fiends for it
Replying to @butter6868
It鈥檚 hard to think big picture while deep in managing agent execution
Have to force myself to zoom out and talk through what鈥檚 valuable to recalibrate on what to exert energy on
Replying to @Abid__Ahasan
Its just a start.. still missing something to help me be more proactive vs. reactive to agent state
It will evolve
Replying to @lassejansen
generally better w/o a doubt
still see unnecessary fallbacks/safe guards, abstractions and type hacking
have struggled to prompt those out as the models are overly defensive (better than the inverse I suppose)
Replying to @catmanyau
Good contracts/verification are key. Resolving merge conflicts is easy, ensuring no regressions, no bad assumptions, no new side-effects is the hard part
Replying to @hassantsyed
I was working on a side project last night on my personal PC. Only have codex access on that machine
Astra and Sol are noticeably worst than Opus and its immediately obvious




