Embodied agents · pre-action validation

The robot that
verifies before
it moves.

Priver predicts the outcome of every interaction, then checks that prediction against live perception — on independent hardware — before a single actuator is allowed to fire.

0
actions on contradicted state
100%
per-action auditability
fail‑safe
default-to-no-motion
Decision path · per interaction
LM
Seed the world model
Object graphs & interactions from a language model
Predict by evaluation
Read a grounded result node — not LLM inference
Validate vs. live perception
Independent circuit, separate clock & power
Gate the actuator command
Motion permitted only on affirmative authorization
Why robots act on bad guesses

A wrong prediction, acted on, can't be undone.

Today's embodied agents commit to physical action on the strength of a world-model prediction — and verify it, if at all, inside the same process that produced it. Shared failure modes get rubber-stamped.

PROBLEM 01

Opaque, un-fixable models

A neural world model spreads each prediction across millions of parameters. You can't inspect why it predicted an outcome, and you can't correct one error without risking every other prediction.

PROBLEM 02

Verification that lies to itself

When the same process that made a prediction also checks it, a stale or mismatched view of the world confirms its own mistake. The check inherits the failure.

PROBLEM 03

Stale state, real consequences

Between the moment a robot perceives and the moment it acts, the world moves. Acting on a prediction the environment no longer supports causes damage, injury, or irreversible change.

The architecture

Predict from a model you can read. Verify on hardware that can't be fooled.

Priver pairs the breadth of a language-model-seeded world model with the fidelity of perception — and refuses to act until an independent structure confirms the prediction holds right now.

Seed

Generate

A language model writes the initial world model: object graphs, states, and interaction outcomes — broad coverage, no training data required.

Ground

Perceive & correct

Sensor data is extracted into a structured scene. Where reality diverges from a prediction, the result node is corrected and marked grounded.

Predict

Evaluate, don't infer

The next event is read from an addressable result node — fast, low-power, offline-capable, and correctable one node at a time.

Validate

Independent check

A separate circuit compares the prediction against live sensor data it receives over its own path. Contradiction = no authorization.

Act

Gated motion

The actuator command reaches the motors only when authorization is affirmatively asserted. Otherwise it's withheld, and the miss is logged.

The validation gate

A hardware veto on every motion.

The gate sits in the command path between the planner and the motors. It runs on its own clock and its own power rail, so a crash, stall, or brown-out in the rest of the robot can't force a movement.

  • Independent failure path. Defaults to no motion — actuation requires an active "yes," never the absence of a "no."
  • Independent perception. Reads current sensor data over its own channel, so it catches a state that changed since the prediction was made.
  • Watchdog & freshness. Authorization decays; a prediction that's gone stale stops being valid before it can do harm.

Try it → flip the live-perception state and watch the command get gated.

priver · validation-circuit · live
Predicted next event
container → RAISEDconf 0.91 · grounded
Live perception
Circuit power rail
CONTEXT-CONSISTENCY
AUTHORIZATION DECISION
ACTUATOR COMMAND
Set the inputs above
Beyond the gate

Caution scaled to consequence.

The same world model lets Priver reason about how risky an action is — and demand proportionally more proof before committing.

Provenance

Per-node audit trail

Every prediction is traceable to its origin — language-model guess or grounded observation — and the full history from one to the other is recorded for supervisory review.

Physics

Constraint validation

An outcome that would push a value past a physical bound or break conservation is caught from its structure alone — before any action, even before the interaction is ever observed.

Corroboration

Multimodal agreement

A single sensor glitch — an occlusion, a dropout — never rewrites a grounded model. Divergence must be confirmed across modalities before it's trusted.

Caution

Graded & staged authorization

Marginal confidence yields a slower, lighter, more reversible action — or a low-energy probe first. Irreversible actions face a higher bar before commitment.

Privacy

Sensor-side boundary

Raw camera imagery never leaves a hardware-enforced privacy boundary. Only the abstracted structured scene crosses into the rest of the system.

Fleet

Federated grounding

When one robot grounds an interaction — a full container topples at the rim — that lesson propagates to the fleet, weighted by reliability, with no raw data shared.

Worked example

Grasp, lift, and the topple it learned to expect.

The language model says "grasping and lifting raises the container." It doesn't know the container is full and gripped at the rim. Priver finds out the way it's designed to — once — then never repeats the mistake.

unexpected outcome PREDICTED · RAISED OBSERVED · TOPPLED FIG · GRASP-AND-LIFT
Step 01 · Seed

The LM's guess

Initial result content predicts a RAISED pose. It doesn't depend on fill level — a gap in the model.

Step 02 · Act

Validated & tried

Live perception matches the predicted state, so the gate authorizes. The robot lifts — and the full container topples and empties.

Step 03 · Ground

Corrected by reality

A new rule is written: full + rim grip + high acceleration → TOPPLED. Marked observational origin, higher confidence.

Step 04 · Transfer

Shared across the fleet

The grounded lesson propagates to similar vessels and to other robots — none of which ever has to topple one to know.

Deploy where action is irreversible

Robots that act only when reality agrees.

Priver is built for manipulators, mobile robots, and autonomous platforms working in proximity to people, under safety-certification obligations, where a known error must be correctable without retraining.