expectedwrong hindsight

OpenAI Bought the App That Lets You Take Over Someone Else's Screen

Multi is gone, its team is inside OpenAI now, and the inference is not subtle.

2 min read 239 words #openai #acquisitions #multi #macos #ai-agents
hindsight — nailed it

openai was clearly building an agent that could use computers. operator/computer-use shipped. the 'you don't acqui-hire a screen-sharing team for slack channels' observation correctly identified the direction.

Multi is dead. OpenAI killed it — gently, as these things go, with a blog post and a thank you — and absorbed the team.

Multi was a macOS screen-sharing tool for people who found Zoom insulting. Low latency, remote cursor control, multiplayer everything, native performance. The kind of product that a certain type of engineer uses and then refuses to use anything else. Small company, excellent team, product that worked.

OpenAI is now on what can only be described as a buying spree — Rockset for the database muscle, now Multi for the real-time collaboration people. You don't acqui-hire a team that built fast screen-sharing infrastructure because you need help with your Slack channels.

The thing OpenAI is clearly building — the thing every one of these acquisitions points toward — is an AI that sits inside your computer and watches. Not a chatbot in a tab. Not a thing you copy-paste into. Something ambient, something that sees what you're doing, something that can take the wheel when you ask it to.

Multi's entire product was: here is a protocol for one person to see and control another person's screen, at low latency, on macOS.

That's not a coincidence. That's a roadmap.

The product is being wound down. Pour one out, or don't — the team is probably fine. But every tool OpenAI absorbs makes the picture a little clearer, and this one was not hard to read.