just had my first 1 hr run ever, with 5.6. insanely good at finding context, spotting error, and iterating on the task until it reaches the goal. truly a good one
happy 5.6 day! testing sol, fantastic. no need to use /goal or any long horizon skill.
the chain of thought reasoning and answers are more verbose too, makes it easy to steer! this a good one
quick review of 5.2 in codex. comparable to 5.1-codex-max!
just as smart and 5.2 is more thorough in planning before execution. this makes 5.2 feel slow. so now its on the user to toggle the reasoning to pick how fast/slow.
Great reflection. Got me thinking on how forward deployed engineers are basically the more technical version of management consultant... The combination of learning a client's tribal knowledge and hooking them into a product makes you not replaceable!
https://t.co/jMpH5VPPLF
@shalahuddinn pecah pecah din!
i used deep research from chatgpt to make the PRD, then had it generate a 5 step plan. then pasted 1 step at a time ke claude code.