Docs · Credentials

ReferenceUpdated 3 weeks ago

Earned by coverage, signed by math

GenBodha credentials are issued automatically from what you actually do: graded labs across your discipline’s skill graph. They are AI-assisted and lab-graded — completion credentials carry no proctored exam or live defense. Each credential is cryptographically signed and verifiable in seconds, and states its own method, so a reviewer always knows exactly what it certifies.

What it is

Evidence that you completed graded, hands-on labs covering a measured share of a discipline’s skills, with AI assistance available throughout.

What it is not

A proctored exam, a timed test, or a substitute for an interview. It certifies coverage and applied practice — not unassisted recall.

Verified (held-out) — beta

Everything above is a completion credential: lab-graded, AI-assisted. Verified is a separate, harder tier — now in early beta. You earn it by passing one held-out checkpoint: a problem the course didn’t walk you through, graded automatically on hidden tests you never see — with the AI pair-programmer switched off in today’s checkpoint workspace (the assistance setting is recorded in the credential). Each Verified credential states exactly what its checkpoint tested — the scope is printed on the credential itself, so it never claims more than the problem you actually passed.

Held out & server-graded — hidden tests grade your solution on our servers; you never see them, and the problem isn’t one the course taught you.
No AI pair-programmer — in today’s beta checkpoints, Claude Code is turned off and the workspace runs network-isolated: no model endpoint is reachable while you work. What never changes: the problem is one the course didn’t teach you, and the tests that grade it stay hidden on our servers. The problem and the grading are sealed.
Private by design — only your checkpoint solution files are captured to grade the attempt, not your whole workspace, and retained only as long as grading needs.
Retry with a fresh variant — not passing just means fix your solution and resubmit after a short cooldown. Each new attempt draws a different problem variant from the held-out bank; if you work through every available variant, that checkpoint pauses until new ones rotate in.

Verified is in early beta: a first checkpoint is gradable today, but no credential is issued and no profile shows a Verified badge yet. When issuance turns on, a badge appears only after a held-out checkpoint passes — signed and verifiable exactly like the credentials above, but certifying that one checkpoint instead of overall coursework coverage.

The ladder

1

Foundational

Shareable now

Passed graded labs covering at least 25% of the discipline's skill graph.

Criteria →
2

Associate

Shareable now

~45% of the skill graph, with at least one passed lab in every skill area.

Criteria →
3

Professional

Shareable now

~65% of the skill graph, with no skill area below 30%.

Criteria →
4

Advanced (Lab Proficiency)

Held — verification pending

80%+ of the skill graph, every area ≥50% — AI-assisted lab work, not a proctored exam.

Criteria →
5

Distinguished

Rare — cross-discipline

Professional or higher across two or more disciplines, with sustained activity.

Criteria →

Tiers are per-discipline (except Distinguished, which is cross-discipline). You can hold different tiers in different disciplines at the same time. “Coverage” is the share of a discipline’s skill graph you’ve reached with passed, distinct graded labs.

Foundational

A graded on-ramp. You’ve passed enough graded labs to cover at least a quarter of the discipline’s skill graph — not yet broad coverage, but a real, signed start.

≥25% skill-graph coverage — passed distinct member labs across the discipline; each distinct passed lab counts once toward the discipline’s skill graph.
Lab-graded — auto-graded by running it: both companion scripts must exit cleanly on your code.

Associate

Demonstrated breadth, not just depth in one corner of the discipline.

~45% skill-graph coverage across the discipline.
Every skill area touched — at least one passed lab in each skill area, so the coverage isn’t concentrated in one place.

Professional

Trained across the discipline, no weak pillars — “broadly trained on the curriculum.” That is exactly, and only, what the signed credential claims: it verifies as Authentic · lab-graded (completion), not as a held-out proof of unaided skill. The artifact never says more than the work behind it.

~65% skill-graph coverage across the discipline.
No skill area below 30% — the breadth floor that separates real coverage from a gameable average.

Advanced (Lab Proficiency)

The top skill-graph tier: near-complete coverage with real breadth in every area. Earned over AI-assisted, lab-graded work — deliberately not called “Expert,” because there is no proctored exam behind it. (The separate Frontier-earned Expert tier — see “Credentials are live” above — is its own path, not this tier.)

≥80% skill-graph coverage across the discipline.
Every skill area ≥50% — strong breadth, no thin spots.
AI-assisted, lab-graded — hands-on lab work with assistance available; not a timed or proctored test.

Distinguished

Breadth across fields. Rare by design.

Professional or higher in 2+ disciplines — two distinct disciplines, held at the same time.
Currently active — sustained recent contribution on the platform: open-source study units, peer review, or published explainers. This is not a lifetime achievement.

How your credential is signed

Every graded lab you pass becomes a signed entry in your evidence record. When your coverage of a discipline’s skill graph crosses a tier threshold, our promotion service automatically issues a W3C Verifiable Credential — which also records its own method (AI-assisted, lab-graded, no live defense) so the claim is never ambiguous.

The credential is signed with an Ed25519 key held in Google Cloud KMS, and the matching public key is published at genbodha.ai/.well-known/did.json. Each credential names the key version that signed it, so the key can rotate without invalidating earlier credentials. Tiers are awarded by an automated rule engine from your evidence — no one hand-assigns a tier. And because every credential carries a public, signed revocation status, a credential we pull (or re-issue) fails verification against our published revocation status from the moment we revoke it — revocability is the feature that keeps the signal trustworthy, not a loophole.

How verifiers check it

A recruiter, ATS, or LinkedIn integration pastes your credential URL into our verifier widget. The check doesn’t ask anyone to take our word for it: the credential JSON, our public signing keys, and the signed revocation status are all public endpoints, and the signature check runs in your browser — not on a server we control. One honest caveat: that check runs through our verifier today, not a generic W3C VC library; standard-library interop is on the roadmap.

  • They fetch our public key once from /.well-known/did.json.
  • They verify the Ed25519 signature in the browser against that key.
  • They check revocation against a public, signed status list, or hit /api/credentials/{id}/verify for a one-shot active/expired/revoked answer.
  • They get a yes/no in seconds — the only calls to us are the public key and the status list, both public, signed endpoints.

If a credential is altered after signing — by you, by us, by anyone in between — the check fails. That’s the point.

Why credentials expire in place

Coverage credentials (Foundational through Advanced) are valid for 24 months. The Frontier-earned Expert and the cross-discipline Distinguished tier are re-checked sooner — 18 months — so the highest claims stay the most current. GenAI skills move fast, so coverage is re-checked rather than assumed permanent — a credential reflects what you can do now, not five years ago.

Where to next

Was this page useful?