I recently made the tough decision to leave OpenAI. It was truly a privilege to make some small contributions to the company’s incredibly massive ambition. I will be forever grateful for @embirico convincing me to join and @thsottiaux’s belief in me. The people at OpenAI were the hardest part to leave behind. I am grateful to have worked directly with literally hundreds and hundreds of the smartest people I’ve ever met — across product, engineering, and GTM. Many of the best people I know also decided to join us during my time there, further proving how magnetic the culture is. Even though I was there for a short time, it feels like the company went through 3 or 4 evolutions. I’m incredibly bullish on OpenAI because it is able to swarm on the hardest problems, move incredibly quickly, and adapt to all the craziness in the world today. I’ve always said, it’s a very non-regrettable choice to work at OpenAI, and that is more true than ever :) Im pretty bad at taking breaks, but I look forward to grinding out a couple days on Maplestory Classic and then getting back to work 🚀

One week since we launched dots! We’ve been working through your feedback and shipped a bunch of improvements: - Faster cloud browsing - Fixed extra phone notifications while you’re already using ChatGPT - Cleaner code blocks and attachments on web - Fewer pauses while waiting for a reply - Faster sidebar loading, smoother animations and less loading jank - Fixed character rendering in Safari and Firefox - Fixed Outlook setup getting stuck - Clearer plan guidance + a way to retry failed activity loads - More stable safety banners on web - Fixed clipped approval buttons and oversized Gmail previews hiding approvals - Fixed enterprise access issues on web and when starting voice - Better dot organization of when to use new versus existing codex threads Up next: - Voice reliability (being improved everyday, but we’ll shout when it’s up to par with our expectations) - Increasing dot’s control of your Codex app: manage project folders, have full Codex thread context, rename threads, and more - Better organization to know what your dot is working on at any given moment - Connect your dot to multiple computers - Better Windows support for local computer control - what else? Thanks for trying our product and giving us lots of feedback. Keep it coming!

We are making OpenAI Dots really good for engineering. There are two ways you can leverage Dots for eng tasks: - connect it to your local computer: this allows dots to control your desktop with CUA, control your Codex app, use the in-app browser, work on your local dev env, and more - connect it to a cloud environment: this is a new product we launched where you can define a cloud dev environment, give it your dependencies, and allow your dot to spin up instances on the fly to do development tasks In either way, this transforms your dot into the orchestrator and control plan, allowing it to delegate to agents working across any environment and prompting those agents forward, similar to how you use Codex today to prompt your codex threads forward. If you’ve tried using Dots for engineering, I’d love to hear your feedback on what is and isn’t working. I’ll also share out improvements as we make them!

Between Muse and Instinct, finding it quite challenging to have a personal agent doing multiple things in a single chat / text thread. Both have been fun products to try and very impressive experiences on a single task though! Curious if anyone had found ways to work around that and have a good experience in a single thread across multiple tasks?

We’re sharing a solution to the Navier-Stokes Millennium Prize Problem, one of the deepest problems at the frontier of mathematics. The proof was produced by a group of agents, using an OpenAI next-generation model significantly more capable than GPT-6 Astra. The problem concerns whether the description of smooth three-dimensional fluid motion modeled by the Navier-Stokes equations can break down. It has remained unresolved for roughly 90 years.

We got a spike in users earlier today because we accidentally went viral on the OSRS subreddit (r/2007scape) 😅 Things that Codex added to the game today: - lots of content moderation lol - Varrock + Falador - A task system for skilling - A dueling system where you can stake (improving) - First 1000 players to get to 50 total level get a party hat

Some interesting stats about the RuneScape clone I launched yesterday ( - 3,000 people created accounts - At our peak, we had 55 concurrent users player and we're averaging about 35 online players per hour - A collective 1,000+ hours have been spent playing the game in just 24 hours - Over 400k experience has been gained by those players - 3,500 monsters have been killed - 3,000 logs cut - 1,000 ores mined - The richest person in the game has 7,000 GP - The highest level skill is 40 woodcutting and the highest total level is 204 Right now Astra is going deep into performance optimizations and then it is going to spend more time expanding the gameplay and content

This is wild I'm currently using the Codex App to play AND build a RuneScape clone with 40 other people. Codex is shipping improvements as we play and I'm taking everyone's requests and adding it in as prompts to Codex as we go lol

In asked GPT-6 Astra in Codex to build RuneScape and deploy it to a ChatGPT Site. You can check out the game here (it is multiplayer, so I'll see you in there!): The most impressive part may be that the demo video below is an Astra one-shot. My prompt: "I want to make a nice 30 second video that will perform well on twitter, show casing the game that we created here - could you make that video?"

Had fun chatting with @MilksandMatcha about all things Codex and Cerebras! Check out the full conversation on Youtube :) These days, I use Codex, Slack, and a browser. Every other tool I use is through Codex plugins or Codex computer use. As more work gets platformed on top of Codex, Ultrafast is going to help everything get done faster!

We're excited to preview Ultrafast mode: GPT-5.6 Sol at up to 14X the speed. It has been fun partnering with the @cerebras team to make it possible to achieve speed without compromising intelligence! We've been using it internally, along with select enterprise customers, and the results are incredible If you want to get notified when we expand access, sign up here:

What $ROOD actually does: - pays for hosted compute: archive reads, fork simulations, traces, receipt publication - bonds the registry: publishing a template or lint rule locks stake, broken code gets disputed, loser burns - reserves publisher handles: pull stake, lose handle. Squatting costs money - opens priority lanes: dedicated fork pool, parallel simulations - funds coverage grants: new rules, undocumented precompiles, adapters. Stakers approve, treasury pays What it doesn't do: gate the core. Lint, scaffolds, conventions run local and free forever. --local cuts cloud out entirely. Token pays for infrastructure and keeps the registry honest. Nothing here promises return.