As of: Draft v0.1 · not accreditation

A teacher in Kabul and an employer in Berlin can both check what you can do, and why.

Sanad is an open protocol for recomputable recognition of skills. Instead of trusting an issuer, a verifier runs a published function over published evidence, with its own rules, and gets a level, a confidence and the reasons. It is a draft, it has no token, and it is not accreditation.

The problem§

Many people can do things they cannot prove. Papers are lost, institutions close, states stop recognising study, and a certificate says little about how it was earned. Employers and universities cannot easily check what stands behind a claim, so they trust brands and borders instead of evidence.

A worked example

An invented learner claims level 4 in a first course in real analysis. The verifier applies the default policy: at least two admissible attestors, one live sitting, attestors with at least three sittings.

Evidence and attestations
ItemKindAttestations
E1Practice runs (replayable)
40 seeded epsilon-delta exercises, replayed from their seeds
replayed from seeds, no attestor needed
E2Live sitting
Live sitting S-2026-014: 90 minutes, two examiners, whiteboard stroke log
Ember (examiner): level 4
Ash (examiner): level 4
E3Reviewed work
Proof portfolio: three written proofs (hash of the PDF)
Linden (reviewer): level 4
E4Peer review
Peer review by a fellow learner
Fern (fellow learner): level 3 ignored (only 1 sitting(s), policy needs 3)

recognised level 4 of 5 · confidence 0.909 · 3 independent attestors · live sitting present

Show the step-by-step explanation (rules R1 to R7)
  1. R1 Checked 4 evidence item(s) and 4 attestation(s): every hash, every reveal against its commitment, and the chain. 0 failed.
  2. R2 Attestor did:sanad:fern (attestation att-04) was ignored: only 1 sitting(s), policy needs 3.
  3. R2 4 attestation(s) and automatic record(s) are admissible on 3 evidence item(s).
  4. R3 Evidence E1 is 1.77 month(s) old: age factor 0.950 (half-life 24 months).
  5. R4 Evidence E1 (auto): levels [3], lower median 3 -> level 3.
  6. R3 Evidence E2 is 0.59 month(s) old: age factor 0.983 (half-life 24 months).
  7. R4 Evidence E2 (sitting): levels [4,4], lower median 4 -> level 4.
  8. R3 Evidence E3 is 1.18 month(s) old: age factor 0.966 (half-life 24 months).
  9. R4 Evidence E3 (artifact): levels [4], lower median 4 -> level 4.
  10. R4 Evidence E4 has no admissible attestation, so it does not count.
  11. R5 3 distinct admissible attestor(s) (did:sanad:ash, did:sanad:ember, did:sanad:linden); policy needs 2 and a live sitting, which is present. Independence is met.
  12. R6 Recognised level = the lower of the claim (4) and the lower median of the live-sitting and artifact evidence levels ([4,4]) = 4.
  13. R7 Confidence = 1 - the product of (1 - weight x age factor x agreement factor) over 4 item(s) = 0.909.

All people and data are invented. The result is computed when this page is built, by the reference code in reference/; node reference/verify.mjs reproduces it. Appendix A of the specification describes the same example.

How it works§

  1. ClaimA person states a skill at a level, for example real analysis at level 4. The skill is a node in an open, versioned concept graph.
  2. EvidenceThe claim is backed by typed evidence: replayable practice runs, a live sitting watched by examiners, reviewed work, peer review.
  3. AttestationAttestors sign what they saw. Marks are committed blind before any is revealed, so that they are independent of one another.
  4. RecognitionAnyone runs the published function with their own policy and gets a level, a confidence and a step-by-step explanation.

Who it is for§

Learners without portable proof

Carry evidence that does not depend on one institution.

Examiners and assessors

Sign what you saw; your record is public and recomputable.

Universities and employers

Apply your own policy and show exactly why a result came out as it did.

Implementers

Build a verifier, issuer, attestor, log or index; run the public vectors.

Funders

A lean plan, a risk register and what support would change.

Press

A fact sheet, and what would show that Sanad is wrong.

Status§

Sanad is a draft (v0.1) for discussion. It has not been reviewed by a standards body, a lawyer or an external cryptographer. It uses decentralised assessment: it does not accredit anyone and does not decide what a verifier must accept. It has no token and no financial component. Until the Open Sanad Foundation (UAE) exists, the author, Dr. Arash Kermani Kolankeh, acts as interim steward; a second steward is being appointed, and the governance charter limits what a single steward may do.

Where to start§