Femi
6.8K posts
The part they aren’t telling you vibe coders: low is indeed fast and good but you need clear guardrails. So I will develop plans using high (or xhigh if it’s a deep architectural thing) and then use high to split the implementation plan into vertical slices which are filed as issues and have detailed instructions and acceptance criteria. Then I’ll tackle all the issues in a loop with test driven development asked on acceptance criteria. When they’re done I do code reviewing and checks to make sure all plan criteria was met. Do smoke testing after that, then PR it where it reviewed by coderabbit. Using @mattpocockuk sandcastle (with a modification for my own work flow practices) and his collection of outstanding engineering skills. He seriously changed the game for me and his ai-hero stuff is WELL worth checking out if you’re wanting to build durable products with LLMs.
really fucking good
Here's how you can start threads from another device (like a Mac mini from your Macbook) 1. settings > connections 2. 'control other devices' 3. 'add device' You should see all the other online devices which you've signed into the Codex App on (in my case, I've added all)
Whoever designed this needs to be fired immediately:
You've been asking for this one... Now in preview: Codex in the ChatGPT mobile app. Start new work, review outputs, steer execution, and approve next steps, all from the ChatGPT mobile app. Codex will keep running on your laptop, Mac mini, or devbox.
