NEW paper from Microsoft Research. If you care about training computer-use agents, this is one to keep. (bookmark it) The team builds 1,000 …

NEW paper from Microsoft Research. If you care about training computer-use agents, this is one to keep. (bookmark it) The team builds 1,000 synthetic computers (each with realistic directory structures, documents, and artifacts) then runs long-horizon simulations on top of them. One agent plays the user and sets productivity goals; another executes the work. Each simulation runs over 8 hours of agent runtime and 2,000+ turns on average, roughly a month of human work compressed into one trace. Training on this experiential data drives significant improvements on both in-domain and out-of-domain productivity evaluations. The framework is explicitly designed to scale to millions or even billions of synthetic user worlds. The bottleneck on computer-use agents has stopped being model capability and become realistic long-horizon training data. This is a credible recipe for generating that data at the scale frontier agents will actually need. Paper: https://arxiv.org/abs/2604.28181 Learn to build effective AI agents in our academy: https://academy.dair.ai/

Source: DAIR.AI (X) | 2026-05-01