Skip to content

Three refusals became verbs - #104

Merged
tannevaled merged 1 commit into
mainfrom
three-refusals-became-verbs
Oct 11, 2026
Merged

tannevaled merged 1 commit into
mainfrom
three-refusals-became-verbs

Conversation

@tannevaled

@tannevaled tannevaled commented Oct 11, 2026 •

Copy link
Copy Markdown
Contributor

divide-pages, remove-blank-pages and repair-pdf are claimed.

covered       76
not covered   38
parity       66.7% (76/114), every claim verified

⛔ Writing down why a thing cannot be claimed is what got it built

Each of the first two sat in this table as not covered, with its reason —
and each time the reason turned out to be the specification:

was written here became
"Poster MAGNIFIES; divide-pages cuts a page into halves that stay half-size" go-pdfkit/ops Divide + pdfops divide
"it needs a blank DETECTOR over rendered pixels, and a detector is a measurement, not a wiring job" BlankPages/DropBlank + pdfops deblank
"no single entry point does the job" (yesterday) DuplexCollate + pdfops duplex

Three in two days. That is an argument for the practice, not just for the
entries.

The detector's measurement changed its own design: the default resolution moved
from 50 DPI to 150, because on A4 at 50 an eight-point full stop covers
exactly one pixel — the same mark a speck of dust makes — so no threshold at
that resolution can tell punctuation from dirt.

repair-pdf is reader's repair, given a name

The claim is the verb, not reader's symbol. The repair itself was always
there and silent: every ops verb has been repairing files for as long as
reader has, so a claim on the symbol would have been true and useless. What
this table can now show is a tool that says so and keeps the result.

⛔ compare-pdfs is not claimed, and the reason is kept

conformance/compare already renders two pages, reduces both to grey, blurs
them and counts the pixels that differ — which is the measurement that tool
needs. But it compares our drawing against poppler's, to judge us; it needs
poppler installed; and it pairs page i with page i rather than
"intelligently". A claim on it would be a claim that we can diff two files a
caller brings, and we cannot. Generalising it is real work, not a table entry.

Measured 2026-10-11 against BentoPDF 3a5f146 (v2.8.8, 114 tools), pdfops
and ops at a376bb2 (= v0.13.0), odf 3d1deee (= v0.7.0), convert
45fbb5d, render 3bb5dcf, app 1feda5d, rtf 6bc4897, markdown
eafebfd, data 128b6c7.

🤖 Generated with Claude Code

divide-pages, remove-blank-pages and repair-pdf are claimed. Each of the first
two was written down in this table as "not covered" WITH ITS REASON, and each
time the reason turned out to be the specification:

  - divide-pages said "Poster MAGNIFIES; divide-pages cuts a page into halves
    that stay half-size". go-pdfkit/ops v0.13.0 has Divide.
  - remove-blank-pages said "it needs a blank DETECTOR over rendered pixels,
    and a detector is a measurement, not a wiring job". It is one now, and the
    measurement moved the default resolution from 50 DPI to 150: at 50 an
    eight-point full stop covers exactly ONE pixel, which is the mark a speck
    of dust makes, so no threshold there can tell punctuation from dirt.

That is three in two days, counting duplex-collate. ⛔ Writing down WHY a thing
cannot be claimed is what got all three built — an argument for the practice
and not just for the entries.

repair-pdf is reader's repair, given a name. The claim is the VERB and not
reader's symbol, because the repair itself was always there and silent: every
ops verb has been repairing files for as long as reader has, and a claim on the
symbol would have been true and useless. What this table can now show is a tool
that SAYS so and keeps the result.

⛔ compare-pdfs is NOT claimed, and the reason is kept. conformance/compare
already renders two pages, reduces both to grey, blurs them and counts the
pixels that differ — which is the measurement that tool needs. But it compares
OUR drawing against poppler's, to judge us; it needs poppler installed, and it
pairs page i with page i rather than "intelligently". A claim on it would be a
claim that we can diff two files a caller brings, and we cannot.

Pasted from the tool's own output:

    covered       76
    not covered   38
    parity       66.7% (76/114), every claim verified

Measured 2026-10-11 against BentoPDF 3a5f146 (v2.8.8, 114 tools), pdfops and
ops at a376bb2 (= v0.13.0), odf 3d1deee (= v0.7.0), convert 45fbb5d
(= v0.4.0 + x/net), render 3bb5dcf, app 1feda5d, rtf 6bc4897, markdown
eafebfd, data 128b6c7.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
@tannevaled
tannevaled merged commit c92060b into main Oct 11, 2026
9 checks passed
@tannevaled
tannevaled deleted the three-refusals-became-verbs branch October 11, 2026 08:48
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant