Synthetic Computers at Scale for Long-Horizon Productivity Simulation
A new methodology called Synthetic Computers at Scale enables the creation of realistic computer environments with folder hierarchies and content-rich artifacts like documents, spreadsheets, and presentations. These synthetic computers are used to run long-horizon simulations where one agent generates productivity objectives specific to a user, requiring multiple professional deliverables over about a month. Another agent acts as the user, navigating the filesystem, coordinating with simulated collaborators, and producing artifacts. The approach aims to scale synthetic data creation for productivity scenarios, addressing the challenge of realistic long-horizon work conditioned on user-specific environments.
Key facts
- Synthetic Computers at Scale is a scalable methodology for creating realistic computer environments.
- Environments include folder hierarchies and content-rich artifacts such as documents, spreadsheets, and presentations.
- Long-horizon simulations involve one agent creating productivity objectives and another agent acting as the user.
- Simulations require multiple professional deliverables and about a month of human work.
- The user agent navigates the filesystem, coordinates with simulated collaborators, and produces professional artifacts.
- The methodology aims to scale synthetic data creation for productivity scenarios.
- Realistic long-horizon productivity work is conditioned on user-specific computer environments.
- The approach addresses the challenge of creating realistic work context stored in directory structures and artifacts.
Entities
—