postinstall is a supply-chain risk and often blocked by companies, CI, or pnpm. Installing agent “skills” should be explicit and auditable: npm can ship the content and manifest, but users should run a clear sync/install command that writes to .claude/skills with hashes, lockfiles, conflict warnings, and uninstall support.
OpenAI be like: Oops, we crashed something. Usage has been reset, @thsottiaux.
Anthropic be like: Your account has been banned... we're unable to reinstate it, @trq212.
Google be like: We dropped a new model… wait, is anyone even watching?
If you ever get tired of managing your Codex threads, just let Codex manage itself! Codex can now create threads, search them, organize them, pin the important ones, and spin up worktrees for parallel tasks.
This is a huge update...
On the recent update to codex, the apps you use in browser stay signed in...
If you watched the latest @lennysan@danshipper podcast you know how significant this is.
Im serious, this could entirely replace web browsers for me.
Instead of having scattered tabs on your computer, you now have them organized by task (thread).
Prediction: Soon, OpenAI will release multiple browser tabs per session, and will learn what tabs you open for what task...
5 Codex features I'm not turning off.
The default instinct is to enable everything you find.
That's how you end up with context soup: random MCP noise, tools you didn't ask for, a model drowning in its own config.
I run 5 flags. Everything else is off.
Here's the setup and why:
My rule for any agent config:
Turn on what improves continuity.
Turn off what sprays noise into the session.
The best setup is not the biggest one.
It's the one that stays quiet until it matters.
I'd rather have 5 clean flags than 15 with three conflicting plugins fighting for the model's attention.
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.
yesterday 2am. deep in a refactor. claude stops responding mid-thought. rate limited.
alt-tab to browser. claude[.]ai. settings. usage. scroll. "resets in 2h3m."
back to terminal. stare at the cursor. three hours. okay.
that was the last time. we vibe-coded a fix.
now my prompt footer just says:
5h[87%@14:30] 7d[75%~2d5h] ex[$19.52/$200]
i see it coming before it hits. i know exactly when it resets. i never leave the terminal.
one bash file. no npm. no daemon. 85 tests because we're not animals.
👇 try and install it to your claude code
paste this into claude code:
Install claude-code-statusline: download https://t.co/g7rDyUMhyM to ~/.claude/statusline.sh, make it executable, and add a statusLine command entry to
~/.claude/settings.json pointing to it with padding 0
or just:
curl -fsSL https://t.co/iD3uUGTXT9 | bash
https://t.co/rU0lKdarYa