JobFrame · The Open Canon
The taxonomy is free. For everyone. For anything.
The Pantone system for jobs: a free, open standard where every job has a number — and computable coordinates.
The only open job taxonomy whose leveling is grounded in measured market evidence — and that shows its work.
The open layer carries the language of work: every cell identifier, the full family/function structure, canonical titles and their variants, the abstract universal-level definitions, and SOC codes where mapped — with every cell's provenance stated. From edition 2026.2 it also carries the canon content — summaries, responsibilities, level guidelines and skills — for the 2,146 cells whose content is corpus-authored. Pay values, computed coordinates, and survey crosswalks are not part of the open layer: the language is free; the intelligence is the product.
2026.2
14,948 cells · 533 families · 243 functions · 44 superfunctions · 7,005 titled entries · 2,109 SOC-mapped cells · 2,146 cells with published content
Attribution (required)
Contains the JobFrame Canon, edition 2026.2, © PeopleAnalyst / jobframe.global, licensed CC BY 4.0.
The trademark rule
The content is CC BY 4.0; the JobFrame name is a trademark. Calling something “JobFrame-aligned” requires conformance to a published edition — name the edition you built against.
What the content layer publishes — and what it withholds
The canon is built from a corpus, not generated. So the open edition publishes the written content — summary, responsibilities, level guidelines, skills — for the 2,146 cells whose content is corpus-authored, and withholds it for the 12,802 cells whose content was written by a model. Those cells still ship their identifier, structure, level, provenance and SOC — only their prose is held back, because publishing model-written text under CC BY 4.0 as “the canon” would not be the same object. Every cell states which case it is in contentBasis, so you never have to infer it from a missing field.
Provenance honesty
Every cell states its provenance: market-priced (a survey job grounds the cell) or model-extrapolated (the cell exists by design). We publish the distinction because you should not have to trust a taxonomy that will not show its work.
What we are not allowed to claim yet
The claims above are not adjectives we chose. Each one is a sentence with a machine-checkable test attached, and this page renders only the sentences whose test currently passes — a over measured artifacts, so the copy cannot get ahead of the evidence. Here are the claims it is refusing us today, and why.
- C2 · withheld
The world's most detailed open job taxonomy: families, focuses, and levels where others stop at the occupation.
- esco: depth not superior to ESCO on: content-depth-below-occupation — ESCO publishes 3043 occupation rows below the ISCO-08 unit group (100% carrying a description, mean 22.24 essential skills), and ALL of it is in its open publication. JobFrame's live v2 canon has 14948 cells clearing the same content bar — but its OPEN edition (2026.2) ships 2146 content rows. CANON-OPEN-CONTENT-1 closed step (a) of the previously-named unblock: descriptions / responsibilities / level guidelines / skills now ship in the open edition, exactly as brief §1 ratified. Step (b) was RULED rather than left open — the 12802 declared model-authored cells do NOT publish their content, because the JobFrame canon is data-first, built FROM the corpus, not generative. So the open edition carries the corpus-authored 2146 rows, and that is the honest ceiling under the doctrine. The remaining gap to 3,043 closes by BUILDING more corpus-authored canon (the SOC march / corpus build-out), not by publishing model prose.
- C1 · withheld
The most comprehensive, thoroughly vetted job taxonomy in existence — open for public use.
- occupation coverage 84/1016 vs O*NET's 1016 — the SOC march is the gap
- vetting predicate waits on TAX; universal-claim comparison waits on the full depth artifact
The rest of the honest numbers
- 84 of 1,016 detailed SOC occupations are covered. That is the gap behind the withheld comprehensiveness claim, and we publish it rather than round it away.
- 16 of 44 superfunctions have market-grounded pay structure; the other 28 abstain rather than guess.
- 2,769 cells are market-priced; 12,179 are model-extrapolated — and each cell states which.
Regenerate the verdict yourself: node --import tsx scripts/build-canon-claim-comparison.mjs — it reads the same fingerprinted artifacts this page does.