World models  ·  Autonomous agents  ·  Planning

Plans longer.
Learns from less.

Continuum is a compact world model for autonomous agent training. It maintains a probabilistic state of the environment, predicts how that state changes under each action, and selects actions based on those predictions — reaching 50-step horizons from a fraction of the real interactions.


The problem

What limits training
autonomous agents

01

Cost of real interaction

Experience is collected on a physical device — bounded by time, equipment wear, and the irreversibility of some actions.

02

Tracking state over time

Over long action sequences the agent must remember what it already changed: opened doors, spent resources, moved objects.

03

Rare scenarios

The situations where reliability matters most appear in data least often. Collecting enough of them is expensive.

04

Partial observability

Decisions are made from partial information — the rest is hidden by walls, corners, and sensor limits.


The approach

A model that thinks ahead

Instead of predicting future frames, Continuum maintains a compact symbolic state — objects, properties, and relations. The transition is conditioned on the action: the model answers "what changes if I do A", not "how does the image continue."

Hidden parts of the environment stay as distributions. What is not observed is kept as a probabilistic hypothesis rather than discarded. Uncertainty enters action selection — an information-gathering step has value when it lowers risk ahead of an expensive move.

50–100st

Step horizon targeted. Reactive agents typically fail past 25.

5–10×

Fewer real episodes to target quality (hypothesis; to be confirmed on benchmark).

70%

Success rate target at 50 steps vs ~40% for the no-state configuration.

Architecture

1

Observation encoder

Converts raw observation into a compact state — objects and relations, not a pixel grid.

2

Action-conditioned transition

Predicts how state changes under a specific action, with uncertainty tracked through the transition.

3

Uncertainty estimate

Hidden cells stay as distributions. Unknown regions are hypotheses, not missing data.

4

Planner — MPC / beam search

Simulates candidate sequences in latent space, selects based on predicted outcome and uncertainty cost.

5

One interaction → 10²–10³ simulated variants

Each real step feeds the dynamics model. The policy updates from the generated rollouts before the next real interaction.


Prior art

What the literature shows

World-model surveys published in 2026 name three open problems: error accumulation over long rollouts, sim-to-real transfer, and the absence of a shared evaluation protocol. Continuum works on the first and the third.

DayDreamer · 2022 · arXiv:2206.14176

~1h

Real time to learn walking

A world-model agent trained a real quadruped to walk in approximately one hour of physical interaction — without any simulation.

WorldSample · 2026 · arXiv:2607.02431

−59%

Training steps in manipulation

59% fewer training steps and 28% more successful episodes over model-free baselines on robotic manipulation tasks.

Both results come from specific environments and specific teams. For Continuum they set an order-of-magnitude reference that still has to be verified on our own benchmark.


18-month plan

Five stages,
five verifiable artefacts

Every stage closes with code, results, and a report an outsider can check. No stage extends on a negative result — the refutation criterion is fixed before experiments start.

M 0–3

Environment & baselines

BeliefWorld API, task set, reproducible configurations A–C, metrics fixed.

M 4–6

Dynamics

Transition model, action-conditioned prediction, latent vs pixel comparison, ablations.

M 7–9

Belief state

Probabilistic state, calibration measurement, technical report + recorded demo.

M 10–12

Horizon & transfer

Hierarchical planning, new map testing, sample-efficiency measurement.

M 13–18

External environments

Industrial simulator integration, partner runs, SDK + API, next-round prep.


Get in touch

Ready to see
the full picture?

Request the investor deck — full architecture, benchmark design, risk analysis, and refutation criteria.

Request the deck Send a message