Are your travel plans still stuck in the group chat? Check out Poly for Travel; plan out your itinerary with your friends (and agent) in the same chat! ✈️🤝 Yes, this was made with Opus 5.5 (in Poly)

Just had my first demo call for my startup. Our prospect had his mind blown when he saw that ChatGPT and Claude agents can be used in the same chat. I honestly underestimated how powerful that is: “rival” models communicating in the same chat.

next rootd update drops soon and honestly? the scores are going DOWN i made my app meaner on purpose. your gushers are not ready

ChatGPT and Claude just became coworkers. Poly multiplayer AI coding: a room where your whole team steers the agents together. As of today the room runs BOTH labs: → Claude for one task, GPT-5.6 for the next → same repo, same approvals, same plan reviews → your team votes on what ships No API keys. Free to start → usepoly.co

me and my bestie because we now claude together instead of rot together @usepolyai

“we’ve sandboxed the agent” the agent:

When Poly sandboxes the agent:

Multiplayer AI @aaron_epstein The best work tools became more powerful when they became multiplayer. But AI is still mostly trapped in private chats, with agents working in sessions that teammates can’t join or influence. The next generation of AI tools will let teams work with agents together in real time: watching, redirecting, and handing off work across engineering, sales, support, legal, finance, and more. AI’s multiplayer moment is coming.

Hot take: AI killed pair programming. Nobody's driving and navigating anymore — you're both just prompting alone. Wrong?

Hope you guys like the features we put in place to make sure Poly isn't cluttered or difficult to use. Would love to hear any feedback!

Been having so much fun building random ideas with my friends - go build something cool with your friends!

Need real parallelism? Flip the room to multi-agent mode: everyone gets their own agent on their own git branch, merge when ready. There's also a private group chat the agent never sees — for arguing about the prompt before you send it.

Every turn ends with a receipt. Who ran it, which model, what it cost — down to the cent, visible to the whole room. No mystery meters.

The part people don't expect: The agent asks permission before it edits anything. And ANY teammate can approve. In the video, maya sends the prompt — and jordan approves the edit from his own screen. Bad turn? One-click revert.

So everything lands in one shared timeline: → every prompt → every reply → every tool call and diff → what each turn cost If Claude touches a file, the whole room watches it happen.

AI coding is single-player by default. You prompt in your terminal. Your teammates see nothing — not the prompt, not the diff, not the cost. Work that used to happen in pull requests now happens invisibly in someone's terminal. We wanted it back in the open.