MOMENTS RESEARCH · 02
Agentic World Models
Internal models that represent state, predict outcomes, simulate alternatives and support closed-loop action.
IN RESEARCH
Active research agenda
Questions guiding our workROLE IN THE MOMENTS LOOP
The system core that connects perception to action by maintaining a grounded, revisable model of people, tasks and environments.
CONTEXT
Text · Image · Audio · Video · Behavior
- 01Perceive
- 02Model
- 03Plan & Act
- 04Assess
- 05Evolve
EVIDENCE
Outcomes · Capability · Uncertainty · Safety
OPEN QUESTIONS
What we are trying to understand.
- 01
What state must an agent retain to act coherently across long horizons?
- 02
How can a model express uncertainty when evidence is incomplete or changes over time?
- 03
How should predicted outcomes be compared before an action reaches the world?
RESEARCH BOUNDARIES
How the work stays grounded.
- Internal models are working representations, not claims of complete world understanding.
- Prediction should preserve uncertainty and evidence lineage.
- High-impact actions require appropriate evaluation and oversight.
PUBLIC RECORD
Methods, results and lessons from the work.
Our research notes will explain what we explored, what the evidence shows and what remains open.
No publications yet
Explore the research questions above while we prepare work to share.