Origin Labs · the platform, beyond agents
One evidence spine. Many domain verifiers.
The agent reference check is the product. Labs is the proof the same evidence contract — proposal → environment → policy → verifier → decision → trace → attestation — generalizes to other autonomous actors. In each case the actor and the domain verifier change; the evidence contract, the deterministic-oracle discipline, and the offline-verifiable attestation do not. These are research demonstrations, not production products.
Verified warehouse
Multi-robot fleet follows the oracle’s BFS-verified plan, avoids a moving human + exclusion zones, and emits a signed attestation. Rendered in 2D and 3D.
Domain verifier: a deterministic grid/collision oracle (finish / escalate / refuse).
Verified operations SLA
Run a multi-wave shift; track utilization, peak-simultaneous, and zero-collision throughput. The fleet earns a signed operations attestation only if it clears deterministic targets.
Domain verifier: a deterministic operations-SLA oracle over the verified run.
Spatial reconstruction verifier Research
A photo/video→floor-plan model proposes geometry; a deterministic verifier separates observed from inferred, voids hallucinated geometry + unbacked metric claims, and escalates under-observed scans.
Domain verifier: a deterministic geometry + provenance oracle. (Private research; not yet a product surface.)
The point of Labs: capability is not permission — for any autonomous actor. An agent action, a robot plan, and a spatial model all earn the same signed, offline-verifiable attestation from the same evidence contract. Demonstrations use synthetic, seeded, replayable data and are labeled as such — reproducible under this verifier, never “safe.”