Spin up a background research agent from a Dot

A user asked their Dot to create a native background research agent, which it set up with read-only web and documentation tools while its computer environments were busy.

What They Did

The user had already occupied the Dot’s own cloud computer and a connected Linux machine with other work, then asked the Dot to create a native background research agent anyway. The Dot said it could, was given a bounded research task with computer and filesystem use explicitly excluded, and reported that this agent was running with read-only web and documentation tools and no computer target assigned.

How It Went

It worked concurrently with the other two running jobs, which led the author to treat it as a separate non-computer worker. The post gives no performance numbers or detailed output quality for that research run.

Worth Knowing

The main catch the author flags is usage: after about a day of testing, they still could not tell how work by the Dot, background agents, launched tasks, and connected machines is counted.