Slips log

The slips log records the register and stance slips caught while drafting the site's pieces, and who caught each one; it holds 72 slips across 8 published pieces and 1 unpublished draft set, 57 of them caught by the same model session that wrote the text.

fieldvalue
statusrunning; the log starts on 2026-09-29
rows72, dated 2026-09-29 to 2026-10-06
drafting model (author column)Claude (Fable 5.1) 48, Claude (Opus 5.5) 16, Claude with model not recorded (author, Language and coordinator sessions) 8
published page/slips/

What counts

A slip is a device or a claim that did not belong in the text: a speech device in silent-read writing, an overclaim, a staged contrast, an undisclosed change. Factual corrections after publication go in each piece's changelog and errata. Drafts that are not published carry the label "(vault, unpublished)".

By who caught it

relationmeaningslips
selfthe same model session that wrote the text57
other sessiona separate model session reviewing it10
hooka mechanical rule at the output boundary blocked the text before anyone read it3
humanStefan1
readersomeone who read the published piece1

The most frequent devices: personification 13, overclaim 9, thing as subject 7, announced count 6, one-line punch 6.

Limits on the page: the log holds only slips someone caught. The "self" count includes checks the model ran under a written procedure (the mode-leak pass), so the self count measures a procedure followed, with unprompted self-correction not counted apart.

Cross-check with the deploy gate

The grader.fixtures check on /conformity/ runs the site's rule table over the "before" text of all 72 slips and catches 0 of them; the 72 are stance and mode-leak devices outside the lexical grader's reach.

Page disagreement: in the claims ledger entry of 2026-10-07, all 72 are counted as caught by a reader (model or human). The log itself has 3 rows with relation "hook", all from a personification rule in the drafting harness (write-scan hook, Stop hook, vestige scanner). The counts here come from the log's relation column.

Files

  • slips/log.csv: the data; columns date, piece, author, device, before, after, caught_by, relation, reader_catch
  • scripts/slips_page.py: builds slips/index.html from the CSV, with the counts by relation, device and piece

Rebuild

From the script docstring, run from the repo root:

python3 scripts/slips_page.py

A new slip is a new CSV row; the page and its counts come from the rebuild.