Hot damn, today was the first time gpt 5.5 actually fixed my typo automatically! So far the experience has been annoying as it has taken my typos in prompts verbatim which is annoying as I'm bit dyslectic.
@cnakazawa This hits so hard, I have kitchen sink package for generic utils in the mono repo and still the agents are creating this one particular function again and again...
@thsottiaux the codex auto approve is misbehaving and blocking actions that I'm explicitly allowing the agent to do, this makes me sad, bit scary to run hours long runs with yolo mode. Tried multiple turns explaining that I approve the actions
@karrisaarinen@linear I've been notion fan, I've build own shared AI notes app (my previous startup), but damn am I happy to be able to run my new company almost fully in linear.
@samlambert Local nvme is in own league. Have once migrated pg from AWS to bare metal to save money and get better performance. Thank god we have metal now
@RhysSullivan 0 skill atrophy, I have data background but spent past few years mainly with product dev, and earlier this year got back to data stuff and some Python skills that I hadn’t used in couple years just came from muscle memory when had to do some interactive stuff. I guess it’s same
@julien_c tmux. I used to run with tabs and iTerm2 before I learned about tmux. I'd love to use ghostty with the native panes, but the shortcuts and custom tmux setup is just too deep in my muscle memory...
We're building new AI lab and all the internal processes are build around @linear there is just one challenge: math rendering. Would be nice to have native math rendering in Linear. @karrisaarinen I bet we're not the only one. Would not want to move the math to some other place
Hardest thing when going back to more math and science driven work is to give my self time to actually read and study the things. Past few years I've just shipped code, and if I'm not constantly shipping I feel that I'm not productive.