OpenReason · How it works

A document you can re-run — and verify.

An ORP document is structured so each part of the reasoning is separately inspectable and auditable. Assumptions are explicit, so it can be forked and re-run; and it can carry a signature, so it can be checked offline by a person or a machine.

A document, in five layers 01
L1
Data provenance
Where did the data come from — and what was excluded or cleaned?
L2
Consequence simulation
What outcomes were modelled, under which assumptions?
L3
Empathy mapping
Who is affected, and were minorities stress-tested?
L4
Accountability
Who decided what, when, and why?
L5
Lineage
What is this derived from, and what forks exist?
Fork and re-run 02

Because assumptions are explicit, anyone can fork a document, change one assumption, and re-run it. Disagreement becomes locatable — you can point to the exact assumption or source where two parties diverge, instead of talking past each other.

Verify · v0.3 03

A document can carry a cryptographic attestation — a detached Ed25519 signature over its canonical form. A reader (a person, or an agent on a rail) verifies it offline: recompute the hash, check the signature against a key it resolves itself, confirm the key hasn’t expired. No need to trust, or even contact, the publisher.

Verification yields a trust tier — self_declared → attested → verified → full — and a claimed tier never exceeds what the signatures prove. For data products, a usage policy then says what may be done, which a rail enforces at activation.

hash ✓→signature ✓→⬡ verified · attested