Vinum ex Machina Casebook: an audited register of 299 AI applications in viticulture, winemaking and the wine business

An audited register of 299 documented applications of artificial intelligence in viticulture, winemaking and the wine business, published in Russian (the original) and English (a translation), field by field. Each record states who runs the application, what the system does, what its source reports came of it and the years the source establishes, and cites at least one source with a resolvable URL — 314 sources in all. Records are classified by domain, technology class, maturity stage, country — every country a case runs in, 42 in all, or international for a worldwide platform whose sources name none — and the confidence grade of the source. Every catalogued record was reopened against its source by seventeen AI agents briefed to find errors rather than confirm claims; the author reviewed the audit's results and its corner cases. Eight of 307 were deleted; of the 299 kept, 181 were fully confirmed and 92 had a discrepancy with their source corrected in the text. No invented source, company or paper was found. The audit's per-record verdicts are not published; the corrected text and a caveat on 85 records (four written or rewritten by the September 2026 re-check) are. This version corrects 21 records (#5, #8, #10, #25, #39, #51, #66, #68, #84, #99, #136, #140, #154, #174, #193, #229, #231, #240, #270, #273, #289) after a re-check of the sources of the keynote this dataset supplements found them wrong or citing a page that no longer resolves. The re-check was carried out by AI agents in September 2026 and the author reviewed its findings; each correction was made from the source and checked against it. Fifteen of the 21 were among the 181 the audit had fully confirmed; the re-check covered only the claims the keynote makes, so the audit's figures above are its own verdicts, and records outside the re-check may carry errors of the same kind. In five of the 21 (#25, #39, #51, #136, #140) the re-check also corrected discrepancies it found in the same sources beyond the claims the keynote makes. What changed is text — names, operators, technology and description texts, results, classification notes, caveats — and sources, in both languages, and the years of one record: two records are renamed and keep their slug, five sources are added and three replaced, two of them by web-archive captures of pages that no longer resolve (314 sources, previously 309), and the number of records with a caveat stays at 85, two added and two removed. No facet of any record changed; the domain, technology, stage, country and confidence counts are those of v2026-09-27. casebook.csv and casebook.json are byte-identical to release v2026-09-29 at vinumexmachina.com/casebook. codebook.md documents every field; methods-note.pdf gives the definition of AI used, how records were collected, the audit, how each record's country was assigned, the re-check and each correction it made, limitations, and the use of AI tools.