Max Stoiber

(since my og post got deleted somehow, reposting) I've joined @OpenAI to help turn ChatGPT into an OS! 🚀 We're building a platform to enable developers to combine apps + intelligence + user context in a way that's never been possible before. ...and we're doing it at unprecedented scale: the iPhone app store had 6M users at launch. The ChatGPT app store had 850M users at launch. 🤯 I'm building a small, high-agency team to work on this. If developer experience and building platforms at massive scale excites you, I'd love to chat; my DMs are open 👋 (3 days/week in SF)

@evilrabbit_ You're going to Austria vs Argentina?! So jealous. Wasn't able to get tickets.

This year we are flying to Phoenix (▲ Offsite), Dallas (WorldCup x2 🇦🇷 matches), New York-Berlin-London-Sydney (▲Ship), Buenos Aires (Family).

@_dmca Happy birthday!

it’s my birthday tomorrow. let’s celebrate. reply with the coolest thing you built with codex. winner gets 3 months of pro.

@evilrabbit_ @instagram @X tbf it got 7x the amount of views on X in that day.

the same post on @instagram has kept going viral for more than four weeks now. On @x, it stopped being shared to other users the day after it was released.

@fat This is what every call booth should look like.

Pierre Computer Company. Sometimes nerd sniped by diffs. Sometimes nerd sniped by AV.

@jxnlco ok but do the 45lb plates say OpenAI though

Me between giving away credits.

@dankuntz Congrats on shipping!! Pre ordered

Batch 1 pre-order is live. We're selling a limited quantity for now, and we likely won't sell these colors / materials again. Pre-order here ➡️ lilguy.net

LILGUY V2 - "STARBOY" 🌟 An adorable, wearable digital pet that behaves like a real living creature. Aluminum / Stainless Steel / Brass / Crystal Batch 1 pre-order ➡️ lilguy.net

@SarahLacard @OpenAI @OpenAIDevs @npew @fidjissimo @dkundel @heccbrent @kagigz Thanks, that's very helpful. This is definitely on our implementation: we don't persist mcp-session-id across turns. I'll forward this interally!

@mxstbr @OpenAI @OpenAIDevs @npew @fidjissimo @dkundel @heccbrent @kagigz Hi there! 👋 Thanks for reaching out - good to meet you! I reproduced the error and recorded a screen capture of the issue flow as well as my diagnosis of the problem. Furthermore, I'll link you to the OpenAI Community Forums thread with more information: community.openai.com/t/bug-chatgpt-…

The longstanding @OpenAI @OpenAIDevs Chat Apps SDK OAuth issues are still by far the largest loss of productive time in my current workflow. @npew @fidjissimo @dkundel @heccbrent @kagigz

@ivanburazin that's insane. i did 1x/month back-and-forth and that trashed me

@mxstbr 1 time each way, at least

I'm flying across continents at least 2x every week. Every time I land, I get IV drips to keep me going. Order it on the plane. Land. Get to hotel. 10 minutes later, nurse at the door. Plug in IV with water, salt, zinc, vitamin C, NAD. NAD almost makes me want to throw up while I'm getting it. But it does the job. Half hour later: good to go! Would die without it haha.

@theo Called it

T3 Code is already at over 4,300 stars on Github. It will surpass UploadThing in a few hours 🤯

@jxnlco @superwhisper Can I get your internal version?! 👀

yo @superwhisper can you support a version of dictation using openai's realtime api w/ web sockets so we can have faster dictation? I really need this I built something internally that works

@JoviDeC 👀 I think this is an issue we're tracking. Will dig in!

Codex on web seems to be fully broken for me since the GitHub outages... It tells me I don't have a GitHub connector, when I go to connectors it shows me a GitHub connection active. When I try to click disconnect it tells me that the link can't be found

@RhysSullivan Hmm interesting. How does this land? "TS samples over the web aren't great, Effect forces you to write great code but is difficult to pick up, code implementation is going to matter less and less, therefore you want your AI to write Effect even if you don't understand it"

this is why you should be using Effect, in the long term it's much more token efficient than writing standard TS your AI gets knowledge about how your whole system works, what errors can happen, and can leverage the o11y to debug it all

TLDR: it is a cron job dispatching tickets from Linear to workers, each of which is a Ralph loop using a Linear comment as draft pad for persisted state. Yes it is all you need. Beautifully designed and minimal. github.com/openai/symphony

@southpolesteve @AlexVianaPro I've resorted to walking up and down stairs at the office because the elevators cut out the wifi signal 😅

Not me walking around downtown Chicago last week with @AlexVianaPro holding my laptop open out to the side like a madman 😬

@aaronglazer I don't even understand

@mxstbr 100B?! Holy shit.

I'm at 2.5B GPT-5.4 xhigh tokens this week and I'm shipping like crazy. I can't fathom 100B tokens 😵‍💫

@ajambrosino @steipete 🫡

One thing @steipete and I talked about over lunch last week was how many former founders are at OpenAI. It’s a really neat part of the culture.

@brian_lovin how did I just unlock new Brian lore that I didn't know!!!

Don’t do this. Half joking, half not, I think experimenting with polyphasic sleep in my 20s broke my brain and I never fully recovered. I can still fall asleep very fast though, which I guess is cool.

@pranaysuyash I had codex write me a quick script that compiles the numbers from my codex history, including pulling it down from my many devboxes. Took one prompt and about five minutes.

@mxstbr how do you track token usage by codex?