Concepts
Physical inference
Oceanir is building models that understand the physical world and help people act on that understanding. Physical inference means drawing a testable conclusion from an observation, retaining the evidence, and using the result to decide what to do next.
Observation → inference → review → next observation
A photo is an observation. A location candidate is an inference. Checking the candidate against independent evidence is review. Asking for another image is a human next step. Each has a different status and should remain distinguishable in the record.
What works today
Orca provides image-based location estimates and visual evidence. The app supports interactive analysis and cases. Platform exposes workspace activity, access, and integration tools. The example workflow connects a real analysis to a review note and another observation.
A recorded request is not an executed action. The example does not dispatch a field worker, send messages, or control devices. Keep human decisions separate from model output.
Where each part belongs
- oceanir.ai: the company, available products, and research.
- platform.oceanir.ai: workspace entry, activity, access, and supported work.
- docs.oceanir.ai: instructions and technical reference.
Start with the platform walkthrough or API quickstart.

