Théo

I just added in my ChatGPT instructions the 2026 state of code. LLMs still thinks we write all of code by hand. Ironic when you're thinking about the fact THEY are writing it now

@vercel Is this considered a harness?

Introducing eve, an agent framework. 𝚊𝚐𝚎𝚗𝚝/ 𝚊𝚐𝚎𝚗𝚝.𝚝𝚜 𝚒𝚗𝚜𝚝𝚛𝚞𝚌𝚝𝚒𝚘𝚗𝚜.𝚖𝚍 𝚝𝚘𝚘𝚕𝚜/ 𝚜𝚔𝚒𝚕𝚕𝚜/ 𝚜𝚊𝚗𝚍𝚋𝚘𝚡/ 𝚜𝚌𝚑𝚎𝚍𝚞𝚕𝚎𝚜/ Like Next.js, for agents. vercel.com/blog/introduci…

@idir_ouhab I meant that it would use the model gpt-realtime-2 and let us speak with Codex in realtime, with screen sharing and it does the agentic stuff in the back with Codex harness Like ChatGPT AVM with video but for Codex

@theo_devs What do you mean by GPT realtime?

What would be something that you wish Codex could do today?

@OpenAIDevs Holy shit thanks, it’s been a while I’ve been waiting for this I also didn’t think we would get it all at once 👀

More of Codex is rolling out across Europe this week. We’re bringing Computer use, the Codex Chrome extension, personalized memory, and Chronicle to Codex users in the EEA, UK, and Switzerland. #codex-2026-06-16-app" target="_blank" rel="nofollow noopener">developers.openai.com/codex/changelo…

@Dimillian @thsottiaux @steipete Looks interesting 👀 Is there any way to listen to it? Video/livestream?

Amazing conversation between @thsottiaux and @steipete about OpenClaw, agentic engineering, and basically building in this new era.

The SpaceX and Cursor merge is interesting, it might get us a 4th player in the AI race with all the data Cursor has and the compute and power SpaceX has Future composer/grok models will be powerful 👀

I’m curious to see the Enterprise numbers of OpenAI with the Anthropic dramas, GPT 5.5, and the fact that they are now available on AWS (which is huge)

Is Anthropic trying to be more open? That’d be very good news

@idir_ouhab Oh okay my bad then

@theo_devs Actually that's the very last release we have You can check the latest updates here: #codex-2026-06-11-app" target="_blank" rel="nofollow noopener">developers.openai.com/codex/changelo…

Big update coming to Codex it seems 👀

@idir_ouhab Ok so that’s the one I’m currently on, I suppose the pushed one has been rollbacked

@theo_devs you should get version 26.609.41114, that's the latest

If you’re waiting for the Codex app for Linux, fill out this form to be updated!

@idir_ouhab I don’t have any update available anymore :/

@theo_devs try to close the app and re-open, that should fix it

@asparks1010100 @ivanjurashere Yes of course but the app itself that wraps up the harness isn’t!

@theo_devs @ivanjurashere Actually ,Codex is open source so you can run other models with the same harness if you wanted to.

@ivanjurashere We never know, was just trying to help haha

@theo_devs I was just joking, man :)

@ivanjurashere No, Codex is OpenAI so that’s only for GPT models

@theo_devs We'll be able to run Fable in it?

@thecsguy Oh okay i don’t know anything about desktop app, that might be why, thanks!

@theo_devs Looks like they rotated their keys which probably forced a full redownload

@reach_vb when does computer use and chrome use comes in EU?

Thursday Codex updates: - Developer Mode for Browser use with controlled CDP access - deeper debugging for network, console, runtime errors, perf, and page state - /init in the app composer - customizable macOS Dock icons - Computer Use for more Enterprise users - Windows per-app access controls - Unread chats in the command menu - Browser use up to 2x faster - rate-limit reset banking for Plus and Pro referral invites to earn more resets during the promotion - clearer summaries, plugin management, usage-limit guidance, and lots of fixes across Browser, automations, SSH, Mobile QR pairing, PRs, MFA, and more

@TamimBinHakim Yes it depends on the tools you use, and some companies invest more time into having proper AI agent guidance such as NextJS, Vercel, Convex, etc I don’t think this issue is resolved

@theo_devs Hey @theo_devs, LLMs often miss the latest docs, which Context7 tried to solve. But Context7 has issues too: it doesn’t always have the exact docs for every package version, and it’s not agentic enough to reliably find or verify the right version-specific context. your thoughts?

I just added in my ChatGPT instructions the 2026 state of code. LLMs still thinks we write all of code by hand. Ironic when you're thinking about the fact THEY are writing it now

@argofowl I fear that, even if OpenAI released Fable 5 level performance, that it would not have the same taste and attention to detail that Claude models are known for

seems like this was a bad benchmark everyone is loving fable 5 from what i’ve seen so far - it’s an outstanding model anthropic cooked imagine if we had fable 5 in the codex app at the price of 5.5 🤪 the future is exciting openai, your move