DegenerateDistributions

@sama can I just get a sticker and a place to buy that bada$$ looking Codex hoodie?!

@techn0eacc @thsottiaux Tell it to ask you questions before working until they have less than x% ambiguity on the scope and intent of your request. I Went through almost 30 questions then it one shorted an entire Mac app for me that only needed very light polishing afterwards.

@thsottiaux I wish Codex would step in with questions when the user wasn't specific enough about their request. Like what'd be the purpose of the app user wants to create.

What are we obviously not getting right with Codex?

@Kevin_Mandla @techn0eacc @thsottiaux @mattpocockuk Read the comment and thought the same.

@techn0eacc @thsottiaux Or you can use the “grill-me” skill by @mattpocockuk on GitHub

Better code review integration with GH - To utilize the GH auto review feature(and separate allotment), which typically takes a few rounds, you have to constantly monitor the PR in GH to know when it’s done. Either integrate into a GH action, have a way to see that it’s complete from the codex app, or a way that running code review in the app counts against the code review allotment that GH does. I am able to move so fast and just “flow” until it’s time for code review. Only way to automate it right now goes against standard usage and not the code review usage.

I have a $50k+ car that I’m in for no more than an hour a day on average with a negative ROI. My main Mac I’m shipping on gets used 8+ hours a day. When you switch up your mindset, spending this much is a lot easier to accept. Lesson: Put your money in what makes you money

@kr0der How much time do you spend in your car each day compared to your computer? Thinking of it that way makes it a no brainer right?

dude a 16 inch M5 Mac with 128 GB RAM can basically buy you a second hand car

@CtrlAltDwayne Agreed. It was fixed after yesterday's release(tested myself for both Chat & Project threads) but seems to have been reverted back to occurring in chat threads in today's Update. Very rarely I'll see it in Projects but is consistently an issue in Chats

I love love the Codex app, but there's a bug that can result in you using more usage than you might realize because sometimes the reasoning effort doesn't change. Even with the latest Codex app version, I still hit this bug where reasoning doesn't change, so I'm using xhigh.

@nicdunz new to actually posting on here but am fully on the Codex train and have started posting daily. Would appreciate the follow!

if you want me to follow you back, comment here and ill have codex go through the comments and review your profiles and follow back everyone who seems worth following

@thsottiaux @romainhuet @Dimillian anyone know if there is a way to start a new thread for an already existing worktree? I've tried both Work Locally then selecting the branch that is checked out on that worktree which errors out, and New Worktree then selected the same branch as above and it spun up a brand new worktree. Am I missing anything?

👀found this little easter egg after the latest update! It pops out to the right of the main thread for quick q&a @thsottiaux @OpenAIDevs

This allows for cookie/cache separation between the worktrees. @theo wasn't this one of your issues with using worktrees a while back? @jullerino idea for T3Code?

integrated portless.sh into my Codex worktrees via bash scripts that run as a Codex Action and it completely opens up the ability to use worktrees for front end work. Highly recommend trying out! Just ask Codex how best to set it up for use with its Worktrees

@LukasKawerau @ajambrosino Put Tailscale on both machines then enable the Codex alpha of SSH in your ~/.codex/config.toml file on your laptop. This setup allows your laptop to use your desktops Codex underlying app server so that history, threads, etc are always in sync

@ajambrosino How do y'all recommend I sync my Codex' history across computers? I have a desktop and a laptop (arch and a macbook) - would be great if they could share insights into how I work where and what's what without me talking to "two different people" as it were

it's still experimental so we hide it a bit, but in the codex app, try: > what have i been doing very inefficiently on my computer (according to Chronicle). make some recommendations. be direct. tell me what i need to hear.

@DevAdventur3s @thsottiaux This is one of the few times I find myself outside of codex on a daily basis. For something that should be so simple to integrate

@thsottiaux Project notes in codex would be nice 👀 Superapp...⚡️

Send us feature requests for codex in the form of an images 2.0 generated image. It makes it easier for codex to implement if we decide to go for it. Saw some good ones today already that codex is cooking on.

@trekedge @jxnlco Amphetamine is your friend! (never thought I’d be saying that but here we are 🤷🏻‍♂️) but seriously when I get the dirty look from the wife while walking around the house like you are on the bike, I turn Amphetamine on and close the lid

@jxnlco Guys will really do anything to avoid getting an ssh devbox

When you gotta bike home from work but your codex needs to finish a task.

It feels so good when you mention a Codex issue to one of the @OpenAIDevs on X, wake up the next morning to a blue in app Update button, and the issue vanishes! You guys really are cooking!! @Dimillian @thsottiaux and all others behind the scenes!

@emanueledpt @vercel Congrats on the success. Just make sure you watch @MatthewBerman ‘s recent video so you don’t make the same costly mistakes as he did!

My first subscription to @vercel After a year of developing saas and websites, the day has finally come I'd see this as a success rather than "losing money"

Get the $20 GPT sub, have 5.5 high review Claude’s last few branches until all P0-2 issues are resolved, read each issue Claude made, you’ll never look back at Claude again. Just be warned the first time you do this you must end up hitting your five hour max bc of the Claude slop!

@thsottiaux @sama Does that mean I should switch from Claude? Claude works great for coding but it just messes with browser use

We will ship again this week. Codex has achieved escape velocity and will keep improving rapidly.

😂😂😂 this guy found the chest code the comm’s. Not just the free shit that comes from it, the way he goes about announcing it is just a *chef’s kiss * Enjoy everyone!

I have a pretty good harness that are good guardrails and keeps the models from veering off path so I’ve had good vibes with medium…EXCEPT for tricky front end work. Had to move up to high to solve those. One I even had to throw back over to Claude last night, which I hadn’t had to do in a while. I base my analysis off of how many code review items surface for the branch and I always classify them into broader classifications. If issues are falling outside of the business logic classification but were able to pass Typecheck and linting, I take note that my reasoning may have been too low and if it makes sense to, I’ll add extra verification into my harness to keep it from happening again. My verifications are all ran as a pre-commit hook so will then be caught going forward ahead of time and fixed

whoever said GPT-5.5 medium is the new standard has clearly never shipped anything serious with it medium drops the ball all the time I've seen this movie before for a long time, people claimed "Sonnet is enough!" when it obviously wasn't people defend the cheaper tear but for me, GPT 5.5 High continues to be the best sorry not sorry