How can we make the cloud computer/browser better for you in dots? We are tracking: - access issues on websites - takeover reliability on mobile - password manager not showing up in 2 step authentication flows - integration with 3rd party password managers like 1Password - latency, especially on starting browser tasks - persistence of your files/apps on the cloud computer - issues with dot controlling an seeing the local in-app browser on desktop Anything else that I’m missing and we should be tracking?

Announcing dots. Dots work 24/7 for you, learn from your feedback, have their own computer, browser and can be connected to over 4k apps in our ecosystem. They’re powered by Astra our best model yet. Included in your Pro plan, without drawing down on any of your usage. You can even call a dot while it’s working. We’re getting you started with your primary dot today and soon you’ll be able to create entire teams of them.

We are so excited to launch dots today, an always on assistant that can do everything. Each dot has its own computer in the cloud so it’s able to take on anything you throw at it! We gave it: - a password manager that syncs your credentials across ChatGPT - browser use and computer use on any native Linux app - local computer use to seamlessly handoff cloud tasks <-> local, so you can work anywhere and it works everywhere Combined with the ecosystem of plugins, and the cloud coding environment, it’s truly remarkable how much it can do. I have used it from simple personal errands like keeping my badminton coach updated with my availability to complex work tasks like monitoring & actioning on product feedback from early testers that involve data analysis, testing website, slack coordination and committing code. The embodiment of the intelligence, the proactive messages, and the full power of a computer make it feel like a true personal assistant that frees you up and multiplies your ability to do ambitious work. The team worked really hard on this one, and we think you’ll really feel the magic when you use it!

we're testing a new sidebar on desktop - v curious for feedback!

With more powerful computer use capabilities in Astra, we are launching new controls & observability that makes it easy for enterprises to safely deploy CUA in their organization. Admins can now: - limit sites & apps that CUA has access to - manage individuals features like uploads/downloads/history/CDP - adjust the amount of human-in-the-loop for tasks by tuning approval settings Every screenshot, action, and approval of CUA are now streamed to OpenTelemetry so admins can easily audit and traceback each trajectory. We've been testing these controls over the last month, and have seen incredible adoption of new workflows across procurement, finance, QA, GTM, marketing, and IT in the world's largest companies. CUA is especially powerful for enterprises because many workflows are done inside legacy/in-house software that do not have any MCPs/APIs. If you are an enterprise interested in deploying computer use, send me a DM and I'll connect you with the team!

I am pleased to see that Ant’s new Claude Code has a version of background computer use on par with the version from Codex from last May this year. Nice work! Shipping great features first turns out to be a great way to encourage other labs to ship too. We will continue to do this until other labs pay more attention to shipping. This is good for everyone and there is a lot of room left to go! We solved computer use in practice for GPT models about four months ago. But computer use is a significantly awesome thing no matter what model you use, and it is important that the industry similarly spends more effort to train their models to be great at computer use, among other elements of model capabilities. As models become more capable and central to businesses and economies, the value of computer use only increases. We should be taking it seriously, and doing the right thing for our customers and the world.

With Astra, ChatGPT is nearly 2x faster at computer use than before. In addition to the amazing new model, we’ve optimized the harness to make computer use run much faster. These optimizations work in existing models too, so tasks are sped up by ~60% in GPT-5.6 Sol. Try it!

This is GPT-6 Astra. Anything you can do on a computer, Astra can do for you. Fast.

The ChatGPT desktop app is now my main browser and my productivity has never been higher

Last week, I pushed a change to the ChatGPT desktop app that makes the loading of threads faster. Not just a little faster. A LOT faster. ✅ Long threads now load over 90% faster. ✅ Decreased the memory footprint of long threads by over 90%.

Two more ships this week for the ChatGPT Desktop in-app browser! Media Tabs: Find your music or meeting without hunting through threads! All tabs with audio/video activity are now grouped in one menu in the side pane. (s/o @ngu_khoi @schwa23) Zero-state autocomplete: Jump back into your browser work faster. Just click the empty URL bar to see your most recently visited pages. (s/o @olivertepman) Let us know what you think, especially if you’re using the IAB as your daily driver. We are still working on bringing extension support safely to the IAB, thanks for your patience!

WebMCP is now supported in the Cloud browser in ChatGPT work! We will bring support to Chrome extension next. S/o to @ndmccormack for pushing hard on getting this in!

It's cool to see the diversity of use cases in the cloud browser over the last 2 days. Just to pick a few, people have used it to: - Update addresses for banks/services after moving to a new place - Renew their car registration with the DMV - Analyze and optimize their ad campaigns - Find affiliate programs for their blog across multiple invite-only networks - Research upcoming podcast sponsors tracked in an internal dashboard - Compare OTC medication across suppliers and purchasing the cheaper option - Reserve backpacking equipment for an upcoming camp trip - Find last minute movie tickets Computer use is so fascinating because it's hard to predict ahead of time how people are going to use it, by the sheer fact that it is a universal connector to the open web.

ChatGPT work's cloud browser + voice is basically Jarvis on your phone!

Today, we gave ChatGPT Work’s computer in the cloud the ability to 1) provide a secure sign-in form to users and 2) persist that login for future sessions. With the ability to login, ChatGPT work can now take care of tasks end to end on any website, instead of just providing information. Over the weekend, while I was in transit to Dolores Park, I had it: 1) order dog food to my in-laws who was taking care of my dog 2) cancel a subscription to my nutritional supplement that I have put off for months 3) choose meals on a meal delivery service that fit me and my wife’s fitness goals 4) book a last minute haircut appointment with my barber This product made it increasingly clear for me that people in the future will work with persistent agents who have their own computer & identity. By giving agents computers, you give them the same interfaces that people have to be able to accomplish any tasks, and it also frees your local devices up from having to stay open all the time. This enables parallelization at scale, and the ability for work to be done anytime, anywhere. However, just giving agents computers and asking people to take over when needed does not make a product experience that is usable and safe for consumers. For this launch, we did a few things: - secure login forms that teleport your credential into the website’s form without the model ever seeing it - auto review model that looks at the form and the destination fields to make sure we don’t make a mistake and the credentials are actually relevant to the task - native integration on iOS (coming soon to android and web) that lets you use your password manager for credentials and messages for 2FA codes - controls to clear cookies on individual sites or across all sites in settings to give users maximum control We still have a long way to go to make the user experience accessible to everyone, and the agent usable on every website for every task, but we’ll get there one step at a time. Let me know what you think!

Today, we are adding support for WebMCP in the ChatGPT desktop in-app browser, and kicking off a hackathon to encourage builders to build agent-native websites. As people and agents work together more, the shape of the open web will also change! Sites will become more agent-native, and serve both people & agents at the site time. By giving agents their own interface, not only can agents use the internet to complete tasks much faster, but we think it can also unlock new user experiences where agents and people can collaborate on the same interface more seamlessly. We'll be doing a livestream at 3pm today on @OpenAIDevs to talk through this future, WebMCP and the hackathon. Come join us!