Skip to research
Time traveling companyExplore

OpenX research

Understanding the world.
Testing what could change.

We explore how world models can help people compare decisions, examine consequences and recognize what they do not yet know.

Explore the research

Selected work

Research concepts, essays and evaluation plans.

SimW1 is presented as a research concept. The linked evaluation plans and illustrative charts are not completed benchmark results or evidence of model readiness.

Research agenda

Better questions.
Testable answers.

What should a world model represent?

People, resources and constraints interact. We study which relationships a model needs to preserve for a specific decision—and which details it can leave out.

What changes when we intervene?

A useful simulation must distinguish observed patterns from assumed causes. We propose tests where an intervention’s effect can be checked against evidence.

When should we trust a result?

Compare against simple baselines, test on held-out observations and report uncertainty. A convincing story is not enough.

Where does the model stop being useful?

Look for failure under new conditions, accumulating error and missing information. Identify when collecting evidence is better than running another simulation.

Our standard

A plausible world
can still be wrong.

Results should be traceable to data, assumptions and an evaluation method. We want to make limitations visible—not hide them behind visual realism.

City image: generated editorial illustration, not a simulation output.