Corpus
The whole archive as one file: every message, its date, its structured fields, its composition score and typing figures with the method version, its set and position, and a tombstone record for anything removed under legal compulsion. JSON Lines, one record per line, UTF-8, with a header record first. The schema is published and carries a version field.
The file is regenerated on every publish. Dated snapshots sit beside it with checksums, so that a citation points at something that does not move. Everything here is in the public domain under CC0 1.0; see how to cite.
Current file
- File
- corpus.jsonl
- Generated
- Entries
- 0
- Sets
- 0 (0 complete)
- Size
- 993 bytes
- SHA-256
4590c1202d0c509b813da14211392207773fd8fa158d502a016a345857bf251c- Schema
- schema.json, version 1.0
- Score method
- none yet (method)
- License
- CC0 1.0
Snapshots
No dated snapshot has been taken yet. Until one exists, cite the current file with its SHA-256 and the generated time above.
Sample skew
The head of the corpus file carries this note, and so does every snapshot:
Writing three messages over weeks to an unnamed future audience selects for people who are online, comfortable writing in English, already thinking about AI, 18 or over, and willing to spend time on something with no reward. This corpus records what a particular slice of people thought at this moment, not what humanity thought. The archive has not yet opened to the public; every entry so far was invited.
Fields
Each entry record carries: an opaque id and URL; the set id and URL, the position (1 to 3), and whether the set is complete; the submission and publication timestamps in UTC; the message exactly as written; the four optional fields (country as ISO 3166-1 alpha-2, occupation category, birth decade, input method), null where not stated; the language detected at build time, marked as detected; the composition score, its method version, and the typing figures; the ids of the other entries in the set; and, for an entry removed under legal compulsion, the date and the legal basis, with the message and everything derived from it null. The schema is authoritative.