We spent 2 hours working in the future

Faster implementation can leave a research team waiting on everything around it. METR explored that possibility in a two-hour tabletop exercise: three researchers imagined working with much more capable agents while their colleagues, external reviewers and experimental infrastructure stayed at their existing capabilities.

In the simulated workflow, understanding results, choosing priorities and obtaining feedback became prominent constraints. Running several ideas at once also created more work to coordinate. These are hypotheses from a role-playing exercise, with assumed agent capabilities and a human game master deciding uncertain outcomes. Its value is in identifying organizational changes worth testing, rather than measuring a productivity gain.