Preview each step's outcome before it runs.
teleon.dev/docs/worldmodel runs retrieval-augmented neuro-symbolic world models (R-WoM) that roll a proposed step forward in 0.3 ms, before it runs, and keep each rollout with the evidence it drew on.
The prediction stayed within limits, so the step went ahead with its rollout on record.
What it does
Retrieval first
Each rollout begins by retrieving related evidence, so its prediction rests on recorded runs.
Learned and symbolic
A learned model predicts the next state, and symbolic rules hold that prediction to known constraints.
0.3 ms per rollout
Rollouts are fast enough to run ahead of every step, and each one is kept with its evidence.
The contract
A rollout request names the task, the proposed step and its starting state; the result is a predicted state, a verdict and its evidence.
R-WoMteleon_extensible_evaluations
Endpoints
| Method | Path | Access | What it does | Status |
|---|---|---|---|---|
POST | /v1/worldmodel/rollouts | RING-2 | Roll out a proposed step before it runs | |
GET | /v1/worldmodel/rollouts/{id} | RING-2 | Read a rollout with its evidence | |
GET | /v1/worldmodel/models | RING-3 | List the world models and their domains |
Listed paths appear in the service’s published files today. This surface comes from the Teleon Omnibus Codex v1.0 to v50.0, Oct 25, 2026. In the one-host plan every path here is served under api.teleon.dev/v1/worldmodel.
For agents
Give your agent the read-only guide first. It reports back before any write.
Words on this page
- World model
- A model that predicts how a system's state changes after an action, used here to preview a step before it runs.
- Rollout
- One simulated run forward from a starting state through a proposed step, used to check the predicted outcome.
- Neuro-symbolic
- Combining learned models with explicit rules and logic, so predictions respect known constraints.