The collection is versioned so downstream users can pin exactly what they built on.
- 6.0 is the dataset edition: the body of records described in
FORMATION.md. - Patch releases (v6.0.1, v6.0.2, …) fix packaging, labels, field hygiene, or documentation. They never add or remove records.
- Minor editions (6.1, …) may add records or fields. Record
ids are stable: once published, an id keeps its meaning in every later edition. - Every release is tagged on GitHub with downloadable, checksummed artifacts — pin a tag, not the branch, if you need reproducibility.
The repository becomes a collection: the founding max-map dataset plus
nine more, each in the same one-file + KEY.md + FORMATION.md pattern.
- Added datasets:
wordnet/(117,791 synsets),conceptnet-human-core/(226,284 human assertions),physical-constants/(415),nutrition/(8,789 foods),sky-and-elements/(12,820: elements, isotopes, stars, planets, moons, Messier objects, constellations),world-factbook/(261 countries + 7,342 cities + 645 rivers and lakes),human-word-norms/(85,889 words + 72,089 associations),measure-of-things/(434 magnitude facts + 233-item eval),akc-eval/(11,250 reasoning items). - Added the What Exists List (
what-exists-list/): a merged catalog of 544,279 things and 1,003,336 associations across every dataset, with an aligned relation vocabulary and a 31,008-link concept↔WordNet bridge. - Data-type audit: fixed World Factbook nested-dict fields, sky
year_discovered(now int/null) and uniform planet keys, 48 physical constants that failed to parse on...,dominant_pos"0"→null, a stray whitespace field, and float consistency inmeasure-of-things. - Added
LICENSE.md(compilation CC BY-SA 4.0; components keep their licenses), a compiler byline, andSHA256SUMS.txtintegrity manifests on the release.
First packaged release.
- Fixed 2 records that had empty labels (
InverseFunctionFn,ProductSerialNumberMappingFn). - Added the
definition_formalfield: formal machine-oriented notation is now separated from the plain-languagedefinition, sentence by sentence (15,416 records affected; no text discarded).max-map.txtshows it as aformal:line. - Regenerated
max-map.vector.jsonlwith naturalized passages (label folded into the first sentence, formal notation excluded from embedded text). - Added
extras/subsets/— four ready-made slices: core-commonsense (32,367), entities (76,755), upper-ontology (2,118), relations (21,810). - Added
extras/wordnet-synsets.csv— 9,026 records mapped to WordNet 2.0 lemma / part-of-speech / sense number. - Record count unchanged: 159,722.