Skip to content
MNSTRY.org
Menu Explore

Topic · 5 articles

Provenance at machine scale

A machine-assembled corpus fails in ways a human-assembled one does not, quietly and at the joins, and the corrections are all forms of writing things down before they are needed. These articles run from our own cleanup ledger through the marking of machine-written material, the proof of deletion, the reconstruction of a decision, and the one sentence that keeps a load-bearing choice from being tidied away.

Start reading The five-minute version

The through-line: The craft and The honest instrument and Cathedral of the mind and Structural, not behavioral , the long reads these articles expand on.

  1. 1 Start where the damage is, in our own corpus, with the joins that every check passed over and the audit that got the size of it wrong by a factor of five. The contaminated corpus
  2. 2 Damage found late is expensive, so the next move is upstream, to what a corpus does with machine-written material at the moment it arrives. Born marked
  3. 3 Marking what you keep raises the harder question of what you can prove you no longer hold. Destroy the key
  4. 4 That erasure works only because the history stays replayable, which is the property the next room takes apart and asks what it actually costs to keep. Reconstruction, not explanation
  5. 5 End at the smallest unit of the same discipline, one sentence written beside a decision so the next hand has to argue with it rather than delete it. Removal conditions

You have walked Provenance at machine scale end to end, from a corpus that could not see its own bad joins to a decision that can still say what would have to be true before it goes. None of it required knowing more than the people who built the thing. It required writing down what they already knew, at the moment they knew it.

Practice Pick one document a machine wrote that you now rely on. Find its seams, the places where two different pieces of work were joined, and read across each one for a topic that quietly changed. Note whether anything about the document marks it as machine-born, and if nothing does, add the mark yourself. Then take its most load-bearing claim and write the one sentence stating what would have to be true before you would drop it. You now know three things about that document you did not know this morning, and the third one will outlive you.

Follow this topic: RSS · agent version