docs: drop em-dashes, state test counts as floors, and lead with today's mechanism - #58
Merged
Merged
Conversation
…y's mechanism
Three documentation fixes the owner asked for, none of which touch code.
**Em-dashes.** README had 49 and SUBMISSION 4. House style avoids them in
anything a person reads, and a judge opens README first. Each one was judged
rather than swapped: a parenthetical pair becomes the commas it was standing
in for, a dash introducing an independent clause becomes a full stop so the
sentence ends instead of growing, a dash introducing a list or a label becomes
a colon, and everything else becomes a comma where the appositive reads
naturally. Two alt texts whose inner list already used commas were
parenthesised instead, because commas there would have produced a run-on.
Deliberately kept: every em-dash inside a quotation of a real artifact string
("ILLUSTRATION - NOT EVIDENCE", "AI-generated illustration - not evidence",
the top banner), and everything inside a code fence. Those strings are burned
into frames and sealed into manifests, so changing them in the docs would make
the docs disagree with the product. Twelve remain in README on that basis,
seven of them in the mermaid diagram; one remains in SUBMISSION.
**Test counts as floors.** README claimed 521 backend and 295 frontend tests
across 31 files; SUBMISSION claimed 521, 295 and 20 real-browser tests. CI run
30846463000 on this commit's parent actually reports 569 backend, 375 frontend
across 35 files, and 38 real-browser specs. That is the fourth drift in a day,
so the fix is not a new set of exact numbers that will be wrong again by
tonight. They are now written as floors ("over 550", "over 350", 38 specs),
the text says they are floors, and both files point at the CI run for the
figures, which cannot go stale because it is the thing that ran.
**The seedream line.** README and SUBMISSION both described a live case
writing "a real seedream still and pixverse clip", with a parenthetical
explaining the still step had since been replaced. Annotated as history, but
it reads as a stale claim because the false-sounding part comes first and a
skimming reader may never reach the correction. Both now lead with what the
pipeline does today, which the code states plainly: still_model is None and
still_source is "schematic:impact_frame", so no model generates a still for
the illustration at all. seedream-5.0-lite is still used, for the sample
scenario photos (scripts/generate_eval_scenarios.py), which is why a sample
case's input attribution names it; that distinction is what makes the
sentence true, so it is kept. The history is one clause at the end, phrased so
it cannot be read as the present.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Documentation only. No code, no assets, no workflow changes.
1. Em-dashes
README had 49,
demo/SUBMISSION.mdhad 4. Each was judged rather than swapped:Kept on purpose: every em-dash inside a quotation of a real artifact string (
ILLUSTRATION - NOT EVIDENCE,AI-generated illustration - not evidence, the top banner) and everything inside a code fence. Those strings are burned into frames and sealed into manifests; changing them in the docs would make the docs disagree with the product. 12 remain in README (7 in the mermaid diagram), 1 in SUBMISSION.Scanned afterwards for what the edit could break: no double commas, stranded punctuation, or headings left mid-clause.
2. Test counts are now floors
README claimed 521 backend and 295 frontend tests across 31 files. SUBMISSION claimed 521 / 295 / 20 real-browser.
Measured from CI run 30846463000 (green on
main): 569 backend, 375 frontend across 35 files, 38 real-browser specs, 75 pen-test.Rather than write four numbers that go stale again, they are stated as floors (over 550 / over 350 / 38 specs), the text says so, and both files point at the CI run for exact figures.
3. The seedream line
Both files said a live case wrote "a real seedream still and pixverse clip", with a parenthetical noting the still step had been replaced. Technically annotated, but the false-sounding claim comes first.
Both now lead with what the code does today:
still_model: None,still_source: "schematic:impact_frame"(src/claimscene/pipeline.py:350), so no model generates a still for the illustration.seedream-5.0-liteis still used, for the sample scenario photos (scripts/generate_eval_scenarios.py:40), which is why a sample case's input attribution names it. That distinction is kept because it is what makes the sentence true. The history is one clause at the end.Matches the live Devpost description.