OpenAI’s always-on agents can now spread across a desk. A single ChatGPT dot can connect to several local computers at once and choose which machine runs each task, OpenAI said on November 9. Until now a dot could lean on only one local machine at a time. The change answers the most repeated request from dot users since the agents launched at the end of September: let one dot work on a laptop and a dedicated machine at the same time.
How it works
Connecting a computer takes a few taps. Open the dot’s profile in the ChatGPT app on that machine, find it under Computers, select Allow, and confirm with Allow access. Repeat the steps on each computer you want to add, and the dot keeps all of them connected.
The dot’s cloud computer stays connected too, so a dot with two local machines effectively has three at its disposal. Each thread still uses one computer at a time, so parallel work means starting a separate thread on each machine.
OpenAI’s examples are about development. A dot can build and test an app on a dedicated machine while the developer keeps writing code on a laptop. It can test a change in Xcode on a second Mac and send screenshots, or run one test on a laptop and a machine with an older SDK and compare the results.
What runs where
A dot can create or continue local Work and Codex tasks on a connected computer, and changing the selected computer does not move an existing task. That matters because dots and Codex are converging: ChatGPT and Codex will share one usage pool in December, so the same agent can move between chatting, coding and running jobs on a user’s machines.
Permissions stay coarse for now. Allowing a computer gives the dot that machine’s local files, code and apps, and OpenAI documents no finer per-folder controls and no way to set different permissions per machine. An offline computer is simply unavailable until it comes back, and it stays connected until access is revoked.
Who gets it
Dots are rolling out. OpenAI lists Pro 100, Pro 200 and Pro 500 accounts for users aged 18 and older outside the European Economic Area, the United Kingdom and Switzerland, plus Business Premium and Enterprise, education and healthcare workspaces in an admin preview. OpenAI doubled Pro limits in late October, and dots are part of what those plans now unlock.
From November 18, access is on by default for the enterprise, education and healthcare workspaces unless the workspace opts out. OpenAI is also building a Business Hub that gives companies analytics and citation controls for ChatGPT, a sign that dots are being aimed at workplaces as much as at individual developers.
Our Read
This is a small change with a large implication. A dot that can reach every machine a developer owns stops being a chatbot with a computer and starts being a coordinator across a desk. The examples OpenAI chose, building on one machine while coding on another, are the daily shape of software work.
The limits are worth noting. No independent test has measured how reliably a dot coordinates two machines, what happens when a selected computer drops offline mid-task, or how the single Allow access step holds up against a machine full of sensitive files. Until those answers exist, the sensible move is to connect only machines you would hand to an agent.
Why It Matters
Dots are OpenAI’s bet that the next interface is an agent with its computer, not a chat window. Multi-computer support makes that bet credible for developers, who live across a laptop, a build machine and a test rig. If dots can run jobs across all three, they become the connective tissue of a developer’s setup, not another tab.
The open question is trust. A dot with access to several machines holds the keys to a developer’s working life, and OpenAI’s documentation concedes the permissions are all-or-nothing per machine. Adoption will depend on whether the convenience of a dot that works everywhere outweighs the risk of an agent that can reach everything.
Background
Dots launched at the end of September as always-on agents inside ChatGPT, each with its own cloud computer and browser, plus the option to connect a local machine. The multi-computer request was the most repeated ask in OpenAI’s developer forum, and the November 9 announcement answers it, with the access widening to business and education workspaces from November 18.
Source: OpenAI’s announcement and Sulat’s report
0 Comments