Open data / release
2026-q4
methodVersion 2.0.0: task scores for statements shared between several occupations are keyed to the occupation as well as the words, and every score row now names the occupation it was scored under.
- Published
- 2026-08-05T16:00:02.007Z
- Model
- claude-opus-5
- Prompt version
- task_scoring_v1.0
- Method version
- 2.0.0
- Sampling
- temperature 0, 3 runs per task
- Schema version
- 1
- US occupations
- 830
- UK unit groups
- 412
- Tasks scored
- 53,309
- UK tasks, UK-native
- 35,043
- UK tasks via crosswalk fallback
- 0
- Previous release
- None, this is the first
- Pipeline commit
- 7a962bb04966f755065625c90bdac02e9e9d79ed
Known gaps in this release
- 471 (task, occupation) placement(s) from the catalogue have no score yet and are omitted from this release rather than published with a zero. They are concentrated in 12 partially scored occupation(s), each of which states the shortfall in its own `gaps` array, and the exposure figures on those pages are computed only from the statements that are scored.
- 47,341 of 53,309 score rows are still keyed on the task text alone. Where that text is shared between occupations the number was produced under one occupation's context and shown for all of them; ledger/OCCUPATION-CONTEXT-PROBE.md measures that at a mean 8.03 points and a 30% band-flip rate on the licensing dimension.
- 4,711 score row(s) have no recorded originating occupation (scoredUnderOccupationId: null). Those pre-date the provenance field and could not be recovered from the batch ledger.
- scores.json is keyed on (taskId, occupationId). Every task id in this release belongs to exactly one occupation (the id carries the occupation), so each of its 53,309 rows is that occupation's own published number and no page renders a row belonging to a different job. What is NOT closed is the scoring context: 12,353 (occupation, task) pair(s) across 1,021 occupation(s) publish a rating that was produced under a DIFFERENT occupation's context, because the task statement's wording is shared between jobs and has only been scored once so far. Each one names the occupation it was measured under in scoredUnderOccupationId, no page presents it as measured for the job on it, and they close as those (text, occupation) pairs are re-scored.
Downloads
Sizes and hashes come from this release’s own manifest. Every file is served from the dated path, with immutable cache headers.
| File | Size | sha256 |
|---|---|---|
| US occupations (CSV)One row per US occupation: SOC code, title, exposure score and band, employment, median pay. | 230.7 KB | fac86cef5258dfee392d0d4b0d94150b807fc45dc07ebdb56c643e14c45b8018 |
| UK occupations (CSV)One row per UK SOC 2020 unit group, same shape, from UK sources. | 129.4 KB | caee16777e2b311f84919607fd99f21f58f298c273b388b298740f14f2c998f1 |
| Task scores (CSV)The full task grain: task text reference, the five dimension ratings, the computed score, the band, and the model’s one-sentence reason. | 21.4 MB | ca6b4b45e194552c635d30ccc45250fa0788e42a8cab9ad8354f2a0f6220c464 |
| Occupations (JSON)The same occupation records this site renders from. There is no private, better dataset. | 4.0 MB | 52e9a162c78ccd1c045dad1dcb76509876cb0a15ead41c096af8dcb71010579d |
| Tasks (JSON)Every task statement, with its source and importance weight. | 21.8 MB | a510336ff8230a85d30b3cbdefa5a3593b43feeb077cbd3deaee0f0839c5aefc |
| Scores (JSON)Every task score, with dimensions, confidence, stability flag and rationale. | 29.1 MB | 0617dc29d84a417217c8f57b9dabc2268635bb9dee1b7da5305199dda8bb0a03 |
| The scoring prompt (Markdown)Verbatim, the prompt that produced every score in this release. | 10.5 KB | 8f9740c0d465409a75ca800a2309be0c65570d9934a17e16376645e0c39b7cb2 |
| Source lock fileExactly which upstream files were fetched, when, with content hashes and licences. | 15.9 KB | aa5bc8e026ab9b401d055f0f35d75e7acb25cf718f5f1bb76684ecb8361a61b7 |
| Diff against the previous releaseWhat moved since last time, and what drove it. | 26.8 KB | 4447d57f13c7614a166455f492c484d7352193afa1e371e42b7f11f86a11864d |
| Diff against the previous release (Markdown)The same diff, written out to be read rather than parsed. | 3.7 KB | fc5e9cefc8a230a3f47b78b66135c625e5e619626513231f0eecbcb3f91c88ef |
Verify a download with shasum -a 256 <file> and compare it to the hash above. The hashes come from the release manifest, which is itself in the release.
Sources this release was built from
| Source | Reference period | Licence | Fetched | Content hash |
|---|---|---|---|---|
| O*NET 30.3 Database (CSV)U.S. Department of Labor, Employment and Training Administration · O*NET 30.3 (verified current at onetcenter.org/database.html) | O*NET 30.3 (verified current at onetcenter.org/database.html) | CC-BY-4.0 | 2026-08-04T00:00:00.000Z | 1a07f893d037ee81083cf6fdcf1c198bedd085beab4597d98702201448876fc4 |
| BLS OEWS May 2025 NationalU.S. Bureau of Labor Statistics · May 2025 estimates (national_M2025_dl.xlsx) | May 2025 estimates (national_M2025_dl.xlsx) | public-domain-usgov | 2026-08-04T00:00:00.000Z | b5855a37f3e03e779f6fbf173d3bbc94aeeff33426aef32fa95ce6d025bab1af |
| BLS Employment Projections 2024-34 (National Employment Matrix)U.S. Bureau of Labor Statistics · 2024-2034 projections | 2024-2034 projections | public-domain-usgov | 2026-08-04T00:00:00.000Z | 46e72b19d9e68475e5b43e74e481cc6b7b8f933a1dcd94f233edca687cce1e30 |
| BLS ISCO-08 <-> SOC 2010 crosswalkU.S. Bureau of Labor Statistics · August 2012 (updated June 2015) | August 2012 (updated June 2015) | public-domain-usgov | 2026-08-04T00:00:00.000Z | 0936c78c4ab2cb0b61d7cfe3dcf0f94d943a4e30b11dade1f11676307e2c0aae |
| BLS SOC 2010 to SOC 2018 crosswalkU.S. Bureau of Labor Statistics · November 2017 | November 2017 | public-domain-usgov | 2026-08-04T00:00:00.000Z | f3a847561562d3e5a30eb848f2902a5f7b02e9c48b3d7f2cc8879899fbc242a7 |
| ONS SOC 2020 Volume 1 - structure and descriptions of unit groupsOffice for National Statistics · 03/12/2025 revision | 03/12/2025 revision | OGL-UK-3.0 | 2026-08-04T00:00:00.000Z | dd29e6b960d8f2d7a390b8842a27feedbe80f8c3e7d524d72a51a9e41fc381ae |
| ONS SOC 2020 Volume 2 - coding indexOffice for National Statistics · Version 13, 03/12/2025 | Version 13, 03/12/2025 | OGL-UK-3.0 | 2026-08-04T00:00:00.000Z | a6e27c94fd18c7fe8bcec7b8ea9afaec7e53c14b4d3c2b7f4c3b205687909fe8 |
| ONS ASHE Table 14, 2025 provisionalOffice for National Statistics · ASHE 2025 provisional (reference April 2025) | ASHE 2025 provisional (reference April 2025) | OGL-UK-3.0 | 2026-08-04T00:00:00.000Z | 85888d690eecc71f062f82b6e75c243f29e2136279cb94dfa4b96d58d03c0d6d |
| Nomis NM_218_1 - APS occupation (SOC2020) by sex by employment typeOffice for National Statistics via Nomis · Apr 2025-Mar 2026 (latest APS 12-month period) | Apr 2025-Mar 2026 (latest APS 12-month period) | OGL-UK-3.0 | 2026-08-04T00:00:00.000Z | c73251b7e4d277b2a6ea90620cb5b1b37483fcd9f10f062ec15e46b08d0b0c92 |
| gaisi-indexHenseke, Davies, Felstead, Gallie, Green & Zhou (arXiv:2507.22748); github.com/drGolo/gaisi-index · commit 0783f0315a5920ce776fc5e83c53fd7f8358f81d; SSC Tasks v0.9.1 (20251124), SSC Mappings ONET v0.9.0 (20251124) | commit 0783f0315a5920ce776fc5e83c53fd7f8358f81d; SSC Tasks v0.9.1 (20251124), SSC Mappings ONET v0.9.0 (20251124) | MIT | 2026-08-04T00:00:00.000Z | The raw payload is a multi-file archive or a git checkout, so no single artefact hash exists; per-file hashes are recorded in data/catalogue/sources.lock. |
| BLS SOC 2018 structureU.S. Bureau of Labor Statistics · November 2017 (final 2018 SOC structure) | November 2017 (final 2018 SOC structure) | public-domain-usgov | 2026-08-04T00:00:00.000Z | ade08af40923266f3a854842e888ca3e93c15b26a147c20a2b12a61f4c4f4077 |
| BLS OEWS occupation profiles (May 2023)U.S. Bureau of Labor Statistics · May 2023 OEWS occupation profiles - the most recent static per-occupation profile pages BLS serves; /oes/current/ and /oes/2024|2025/may/ now redirect to the interactive tables | May 2023 OEWS occupation profiles - the most recent static per-occupation profile pages BLS serves; /oes/current/ and /oes/2024|2025/may/ now redirect to the interactive tables | public-domain-usgov | 2026-08-04T00:00:00.000Z | The raw payload is a multi-file archive or a git checkout, so no single artefact hash exists; per-file hashes are recorded in data/catalogue/sources.lock. |
| O*NET-SOC 2019 to 2018 SOC crosswalkNational Center for O*NET Development · O*NET-SOC 2019 (1,016 rows) | O*NET-SOC 2019 (1,016 rows) | CC-BY-4.0 | 2026-08-04T00:00:00.000Z | 506fff0a3ca0d26137b9389322413022a4348bead559c4dd24edeb2cea86aa89 |
| Collab365 Futureproof task-scoring promptCollab365 · task_scoring_v1.0 | task_scoring_v1.0 | CC-BY-4.0 | 2026-08-04T00:00:00.000Z | The prompt ships verbatim in this release as prompt.md and is hashed there in manifest.files. |
| Anthropic Claude Opus 5Anthropic · claude-opus-5 | claude-opus-5 | proprietary | 2026-08-04T00:00:00.000Z | A hosted model has no fetchable artefact to hash; the exact model id is the pin, and it is a component of every score cache key. |
Cite this release
Cite the release, not the site. Release 2026-q4 will still say exactly what it says today, at the same address, for as long as this project exists.
Plain text
Collab365 (2026). Collab365 Futureproof: task-level AI exposure for US and UK occupations, release 2026-q4 (methodVersion 2.0.0, promptVersion task_scoring_v1.0). https://futureproof.collab365.com/data/2026-q4. Licensed CC BY 4.0. Built with O*NET data (USDOL/ETA, CC BY 4.0); ONS data (Open Government Licence v3.0); GAISI task framework (arXiv:2507.22748, MIT); BLS data (public domain).BibTeX
@misc{collab365futureproof2026q4,
title = {Collab365 Futureproof: task-level AI exposure for US and UK occupations, release 2026-q4},
author = {{Collab365}},
year = {2026},
url = {https://futureproof.collab365.com/data/2026-q4},
note = {Release 2026-q4, methodVersion 2.0.0, promptVersion task_scoring_v1.0, CC BY 4.0}
}