Artificial Knowledge Collection — v1.0.0

The Artificial Knowledge Collection becomes a collection: the founding common-sense map plus nine more human-authored or human-measured datasets, and one merged file that ties them all together.

⭐ Top download — the What Exists List

What-Exists-List.jsonl.gz is the merged master file: 544,279 things that exist — concepts, word senses, words, elements, isotopes, stars, planets, countries, cities, foods, physical constants, everyday things — in one unified schema, connected by 1,003,336 associations. One aligned relation vocabulary lets a single chain cross from the hand-built ontology into WordNet into the everyday-assertion graph without a translation table; a 31,008-link bridge joins the concept map to WordNet directly. Start with What-Exists-List.sample.jsonl (readable) and What-Exists-List.manifest.json (counts).

Individual datasets

download	what it is
artificial-knowledge-collection-6.0.jsonl.gz	the founding 159,722-record common-sense map
wordnet.jsonl.gz	Princeton WordNet 3.1, 117,791 synsets
conceptnet-human-core.jsonl.gz	226,284 human-typed everyday assertions
physical-constants.jsonl.gz	CODATA 2022 constants + SI prefixes + conversions
nutrition.jsonl.gz	USDA SR28, 8,789 foods, 679,045 measured values
sky-and-elements.jsonl.gz	118 elements, 3,352 isotopes, 9,096 stars, 13 planets, 46 moons, 107 Messier objects, 88 constellations
world-factbook.jsonl.gz	261 countries and territories
world-cities.jsonl.gz	7,342 world cities with coordinates and population
world-physical-features.jsonl.gz	645 named rivers and lakes
human-word-norms.jsonl.gz + .free-associations.jsonl.gz	85,889 rated words + 72,089 association pairs
measure-of-things.jsonl.gz	434 everyday-magnitude facts with ranges
akc-eval.jsonl.gz	11,250 provable yes/no reasoning items

Everything in one file

artificial-knowledge-collection.master.tar.gz — the entire collection (all datasets, docs, and the What Exists List) in a single archive.

Verify your download

SHA256SUMS.txt covers every asset:

shasum -a 256 -c SHA256SUMS.txt --ignore-missing

Compiled by Max Freedom Pollard. The compilation and the datasets original to it (What Exists List, The Measure of Things, AKC-Eval) are under CC BY-SA 4.0; each component dataset keeps the license of its source. See LICENSE.md.

Artificial Knowledge Collection v6.0.0

First packaged release: 159,722 records (137,912 concepts + 21,810 relationship types), 355,056 taxonomic links, 33,984 domain/range constraints — every record hand-written and human-checked.

Downloads

asset	what it is
max-map.jsonl.gz	the whole dataset, one JSON record per line (universal, no deps)
max-map.parquet	compact columnar format for pandas / Polars / DuckDB
max-map.txt.gz	plain human-readable text, one concept per block
max-map.vector.jsonl.gz	one embed-ready passage per concept, for vector memories and agents
edges.csv.gz / edges.parquet	the 355,056 links as subject, relation, object triples
core-commonsense.parquet	slice: 32,367 defined general concepts
entities.parquet	slice: 76,755 named individuals
upper-ontology.parquet	slice: 2,118 metaclasses
relations.parquet	slice: 21,810 relationship types
wordnet-synsets.csv.gz	9,026 records mapped to WordNet 2.0 lemma/POS/sense
SHA256SUMS.txt	checksums for everything above

Verify a download: shasum -a 256 -c SHA256SUMS.txt --ignore-missing

What changed in this packaging (vs. the raw repo history)

• New definition_formal field — formal notation separated from plain-language definitions (15,416 records; no text discarded)
• 2 empty labels repaired
• Naturalized vector passages
• New subsets + WordNet crosswalk

Full details in VERSIONS.md, field reference in KEY.md.