The over-use profile — what Ink over-uses, found by counting rather than by reading measure · built into the pick · 2026-09-13

The first thing every business with a result did (the industry and the habits §7) was to count what its writer over-uses against human writing, and only then decide what to fix. This page is that count for Ink: every English piece the box has published, against the bank of the majors' essays. It runs by python exam/ink_profile.py, no model call anywhere, and it is re-run after every change to the writer. 142 English pieces · 118,971 words · 2026-08-28 → 2026-09-13 · against 341 bank pieces · 447,151 words · run 2026-09-13 23:33

1 · How it counts

Short answer: for every word and phrase, how often Ink uses it against how often the majors do, and only the ones that come back across many pieces and many mornings.

2 · The words

Short answer: three families. The verdict frame's own words (cannot, does, whether, nobody, merely, remains); the wise-line vocabulary (thing, real, whole, question, line, sentence, machine); and the beat's words (price, market, report, ten), which are the news, not the writer, and are left in the table so the reader can see the difference.

No word in the first version is banned anywhere, and no ban list would catch it. What the count sees is the pile-up: that is the, whole, nobody, whether, will not, simply are each several times over the majors' rate, and they arrive together.

3 · The phrases

Short answer: the same frame, now visible as phrases. That is a runs at 35 times the majors' rate, that is the at 29, that is not at 23, is the one at 22. The over-used four-word phrases are only four, and three of them are the frame again.

4 · The last sentence of a paragraph

Short answer: our top habit is real and it is a shape, not a length. Ink's paragraph-final sentences are no shorter than the majors'. They open on That is / This is / It is four times as often, and they are short-and-built-on-"is" two and a half times as often.

paragraphs with two or more sentences: ink 1,317 · bank 5,491

Two things the table settles. First, the majors close paragraphs on short sentences as often as Ink does (a quarter of the time on both sides), so "never write a short sentence" was aimed at the wrong thing. Second, the gavel is positional in Ink and not in the bank: the majors open a sentence on That is at the same low rate wherever it sits; Ink does it twice as often at the end of a paragraph as elsewhere. No outside list counts position, which is why yesterday's research found this habit nowhere.

The phrases that over-populate Ink's last sentences, against the bank's last sentences:

5 · One morning's pieces against each other

Short answer: by exact phrase, a morning's pieces share almost nothing, and that measure cannot see the sameness readers feel. By shape they can be seen: in the Flash era, pieces from the same morning are more alike than pieces from different mornings.

The phrase column is a null result to keep: the bank's own number is inflated by outlet boilerplate (a paywall line, a byline pattern), and even so Ink's mornings sit below it. The shape column is the finding. Every V4 Pro morning sits at or below the bank's likeness; four of the five Flash mornings sit above it, and above Ink's own cross-morning baseline, with 09-09 and 09-12 the most alike. Something set per morning — the shape menu, the brief's accent, the same four-draft pick — makes that morning's pieces rhyme. This is the number yesterday's research said we lacked (item 6), and it is now in the script.

6 · By era — the same phrases, read against each writer version

Short answer: the verdict frame did not go down when the never-lines went in. On the 25 pieces of era C it is higher than in either earlier era, on most of its phrases.

7 · What changes