Key Takeaways

  • OpenAI provisions each Dots assistant with its own dedicated Linux virtual machine in the cloud, bypassing the limits of simple browser scraping.
  • Giving agents full desktop OS environments lets them use existing legacy software built for human interfaces without custom API integrations.
  • In one personal test, Ari Weinstein cut a tedious two-hour custom grocery order to 15 minutes by delegating it to Computer Use.
  • A primary real-world application of Computer Use is closing the development loop by having agents test and click through the software they write.

Dedicated Linux Computers Change What Agents Touch

Most agent automation runs into a hard ceiling: the browser sandbox. If a tool lacks a clean API or lives inside a legacy desktop application, standard web scrapers fail. OpenAI took a different path with Dots.

“Traditionally we've have access to a browser in the cloud or it has access to your own computer, but now you get your own entire Linux computer in the cloud and so it can run full desktop applications,” Weinstein explained. “Dots are really cool product because each dot has access to its own Linux virtual computer in the cloud.”

This architectural shift matters because human software was never built for API-first consumption. “All the software in the world was designed for humans and now agents can use that same software and you can delegate to the agent,” Weinstein said. By running on a persistent cloud virtual machine, the agent does not require users to keep their own laptops open or maintain local scripts. You assign a multi-step task and walk away.

Weinstein shared a concrete example from his daily life: placing a strict grocery order with exact macronutrient requirements. “I can say like, I want this many grams of chicken and this many grams of of of rice. Um, but it was so complicated, it took me two hours to do an order. And I found that I could ask computer use to do it for me and it did it in 15 minutes.”

Automated UI Testing Closes the Loop

Beyond consumer chores, persistent virtual machines unlock a major developer workflow: agents that test their own code. Until now, coding agents wrote functions, ran unit tests in a terminal, and stopped. They could not verify whether the user interface actually worked.

“One of my favorite use cases for computer use actually and one that we see a lot in the wild is computer use letting the agent actually test the software that the agent has built which is far more consequential than it sounds,” Weinstein noted.

When an agent controls a full desktop operating system, it can launch the local development server, open a browser window, click buttons, fill forms, and catch visual bugs. It inspects what real users see. That tight feedback loop turns code generation from a passive suggestion engine into an autonomous builder that verifies its output before handing it over to a human developer.

What to Do With This

Audit your team's manual operations this week and find the single workflow that requires jumping between three desktop apps or internal portals without APIs. Run a test script against an agent in a sandbox virtual machine to execute that flow end-to-end, measuring both completion time and error rates.