Six tiers. One ledger. No promotion without proof.

Most platforms in the international-student space sell L0 dressed as L4. Albi never does. Every entry on a public Evidence Portfolio has its tier visible, the artifact linked, and the reviewer named.

Why six tiers and not three

Three tiers (self / verified / public) collapse the place where students get hurt: the gap between “an artifact exists” and “an independent reader can use it.” That gap is where ghostwritten essays live, where padded resumes live, where unverified internships live. Six tiers force us to name the gap explicitly.

The shape mirrors what an admissions committee or PI actually does: read everything as L0 until they can find something at L3+ to anchor their judgement. We make that the visible floor, not the hidden one.

The tier table

L0 Self-reported
Student logs the work.A timestamped self-report. No external trace yet.
"I read three papers on EGFR resistance this week."Honesty floor only. Cannot be cited externally.
L1 Artifact exists
A file, note, or output exists.A reading log, annotation set, code stub, or notebook in version control.
A Notion or Markdown file with annotations, dated.Proves time was spent. Does not prove it was useful.
L2 Reviewed by project owner
Internal reviewer confirms artifact has substance.A mentor or m563-assigned reviewer leaves a signed note that the artifact is non-trivial and original.
Mentor: "Yes, the annotations on Robles 2023 are correct and complete."First external eye. Still inside the relationship.
L3 Commit / source linked
Linked to a public commit, PR, archived URL, or DOI.The work has a referenceable address: a GitHub commit, an OSF DOI, a Zenodo record, a published preprint.
PR #142 merged into the m610-612 dataset, commit a7c9f01.First independently verifiable claim. Threshold for paid Evidence Atlas.
L4 Used by project / publicly visible
Downstream consumes the work.Another project or publication cites, uses, or builds on the artifact.
m590 knowledge graph imports the student's anchored claim record.Work is no longer just yours. The ecosystem treats it as a building block.
L5 Recommendation eligible
An independent scholar has enough to write an evidence-backed letter.Multi-quarter L3+ trail with named reviewers, repeated artifacts, and at least one L4 dependency in the wild.
Six months of merged contributions to m590 + two L4 dependencies + named co-reviewer.We stop at supplying the evidence packet. The recommendation itself belongs to the scholar (Charter §3.8).

How promotion happens

Promotion is event-driven, not time-driven. A piece of work moves up a tier when a new fact becomes true:

  • L0 → L1 — an artifact is committed.
  • L1 → L2 — an internal reviewer signs off.
  • L2 → L3 — the artifact gets a publicly addressable handle (commit hash, DOI, URL).
  • L3 → L4 — a downstream project depends on the artifact in production.
  • L4 → L5 — a sustained L3+ record exists with named, blinded reviewers and at least one independent L4 use.

A demotion can also happen — if an artifact disappears, becomes unverifiable, or is withdrawn by the student under Charter §5 consent revocation.

Payment firewall (Charter §3.16)

Payment buys structure, supervision, and access to coaching slots. It does not buy a higher tier. The reviewer is blinded to:

  • Whether the student is on a paid plan.
  • Which tier they are on.
  • Whether their renewal is upcoming.

If the system-wide decline rate falls below 30%, an external audit (m562) fires the next monthly cycle. That floor is the protection against payment-induced inflation.

The evidence packet

When a student is at L5, Albi prepares an evidence packet — not a recommendation. An independent scholar reviews the packet and decides, on their own authority, whether to write a recommendation letter. Albi never sells, packages, or promises recommendations.

What an evidence packet contains:

  • The append-only contribution ledger, time-stamped.
  • Each artifact at L3+, with the public address.
  • The named reviewers, with their decline-rate history and conflicts of interest.
  • The redline grep history — i.e., a public record that the work itself never tripped a Charter forbidden phrase.

What an evidence packet does not contain:

  • Ghostwritten essays.
  • Outcome guarantees.
  • Anything paid-for that pretends to be earned.

Charter §3.11 in full