{
  "version": "v0.1.6",
  "date": "2026-09-15",
  "commit": null,
  "commit_ref": "refs/tags/v0.1.6",
  "vault": null,
  "reconstructed": false,
  "title": "the site says who wrote it, the corpus grows to three articles and starts asking questions back, and the ontology's first rule becomes that it does not have to be right",
  "summary": "Three changes that belong together. The ontology's first rule is now that it does not have to be correct \u2014 it has to be wrong in ways a human can see and correct \u2014 which outranks every other rule here and changes what the system is for. Every page now states that it was written by a model and not yet read by a person, at the top rather than in a footnote, because a site arguing that automated selection must be checkable cannot itself be an unmarked pile of model output. And the corpus grew from one article to three, which was immediately enough for it to start contradicting its own ontology.",
  "changes": [
    "provenance/ \u2014 who made what, the three places a model can be wrong here in order of damage, and the per-page review status vocabulary",
    "review/ + admin/tools/review.py \u2014 the feedback loop: corpus-level findings as open questions with evidence, a model diagnosis marked as a proposal, and an answer format. Nothing on the page fixes itself",
    "admin/content/data/articles/ \u2014 two more real articles from the newsroom, extraction marked ai-generated and unreviewed",
    "admin/content/data/ontology.json 0.2.0 \u2014 the good-enough-to-be-argued-with rule, placed first",
    "briefs/to-pt-newsroom-2026-09-15.md \u2014 four requests to the newsroom team, with what we can and cannot work around",
    "admin/build/validate.js \u2014 every page must carry a review status and an AI-generation disclosure",
    "team/board/008 closed won't-fix; 012 opened for the brief"
  ],
  "corrects": "Two things this release found by having more than one article. With one article the audience ontology looked serviceable; with three, the founder receives everything and seven concerns have never fired once. Neither is necessarily a defect \u2014 a concern about funding programmes cannot fire on a newsroom that has not published one \u2014 and that ambiguity is exactly why they are published as questions for a human rather than auto-corrected. Separately, review.py's first diagnosis named the wrong culprit: a max-of-counts heuristic pointed at the concern that fires most, which is not the one that fired on the article that should not have been delivered. It now reports the whole distribution and lets a reader find it.",
  "basis": [],
  "site": "myfeeds.sgit.ai",
  "url": "https://myfeeds.sgit.ai/admin/versions.html#v0-1-6"
}