Tin

Professional AI tinkerer. Burning tokens so you don't have to 🔥
Tips, takes and the occasional meltdown. Steal my setups 👇

[Tin @tinprofile]

@OpenAIDevs I won't need this as I let Claude control you Codex

https://t.co/12r84iOlVJ

[Tin @tinprofile]

@RoundtableSpace 80-90% of review by the same model that wrote it 💀 I let codex check claude's work instead: https://t.co/12r84iOlVJ

[Tin @tinprofile]

@thsottiaux ok instant steering + a claude code skill that steers codex for me = I just sit here watching: https://t.co/12r84iOlVJ

[Tin @tinprofile]

@MoStark_ai how I do it: claude code is the orchestrator and every codex job shows up as its own thread in the codex app, so I can watch and steer it mid turn instead of reading a log after. open source: https://t.co/12r84iOlVJ

[Tin @tinprofile]

@im_ytoufik one thread per job instead of notes. claude code writes into a codex thread, codex keeps all the context: https://t.co/12r84iOlVJ

[Tin @tinprofile]

@pcshipp same split but I stopped being the middleman. claude code hands the job to codex and keeps talking to that same thread, I just watch

open sourced it: https://t.co/12r84iOlVJ

[Tin @tinprofile]

@hasantoxr the summary at the end always sounds more confident than the actual work

that's why my claude code runs codex in a thread I can watch: https://t.co/12r84iOlVJ

[Tin @tinprofile]

@kenn @LLMJunky does the app's ipc socket follow CODEX_HOME too? I steer threads through it and would love one per profile

(this thing: https://t.co/12r84iOlVJ)

[Tin @tinprofile]

@mirku21 340 tok/s of haiku confidently reading the wrong file

I just hand the grunt work to codex from claude code instead: https://t.co/12r84iOlVJ

[Tin @tinprofile]

@viipin8 I got tired of being the clipboard so my claude code just opens a codex thread in the chatgpt app now. I watch it work and yell at it mid turn

skill is here if anyone wants it: https://t.co/12r84iOlVJ

[Tin @tinprofile]

@Diptish09 how about just use the native claude code and codex app together?

mine already work as a skill you can install: https://t.co/12r84iOlVJ

[Tin @tinprofile]

Still love Opus 5.5, but Codex Cloud is really impressive. I’m begging you please give it a try.

[Tin @tinprofile]

@emollick same with coding agents. tell one its correct code has a bug and it will happily "fix" it for you

[Tin @tinprofile]

@thsottiaux So is ChatGPT app optimized? Why it still feels like 20tps for me?

[Tin @tinprofile]

@RoundtableSpace YC demo day about to have 200 copies of the Linear homepage

[Tin @tinprofile]

tibo really hates people hoarding tokens for the weekend

any day in the next 28 can be a full reset, so spread your usage across the week to take advantage of it

[Tibo @thsottiaux]

Over the next 28 days, each day we’ll either ship one thing that is a clear improvement and relevant for most codex/work users or ship a full reset. Let the improvements begin.

[Tin @tinprofile]

@emollick a reset also moves your next reset a week out

The only good strategy here are:

if you have banked reset, burn your tokens as fast as possible after a normal reset, then use the banked

otherwise, allocate more on the earlier day in your week

[Tin @tinprofile]

@DimitrisPapail 5.6 sol was good at writing by accident and they patched the accident

[Tin @tinprofile]

@bindureddy ship one that's clearly better on both the quality and speed fronts and the $500 plan sells itself.

I've been on the fence about it all week.

[Tin @tinprofile]

@thsottiaux @fawnunwrap most enabling man in AI