Skip to content

docs(blog-site): announce archive links saving the original article - #1190

Open
FagnerMartinsBrack wants to merge 3 commits into
mainfrom
rp/youthful-ramanujan-m4a1as
Open

FagnerMartinsBrack wants to merge 3 commits into
mainfrom
rp/youthful-ramanujan-m4a1as

Conversation

@FagnerMartinsBrack

Copy link
Copy Markdown
Member

What

One new blog post, save-wayback-machine-and-archive-today-links, drafted from the commits that landed on main after the Gmail integration post (2026-10-03): the archive capture cluster (4a1f1e5, a1a9ea7, c3d479a).

The post announces that a Wayback Machine or archive.today link now saves the article it preserves: the save files under the original address, the capture is judged against the live crawl instead of replacing it, and a gone origin falls back to the capture.

Facts checked against the source

  • The 49-word calendar-page save comes from the 4a1f1e5 commit message.
  • The 7 archive.today hosts are the list in wrapper-url.ts.
  • Tie keeps the live crawl, capture can't win on media, and the dead-origin fallback (failed / not-found / blocked, saves and recrawls) come from the .architecture/2026-10-04-a1a9ea793 snapshot.

Editorial decisions

  • Structural variety lookback ran over the last 8 published posts; the opening (numeric), TL;DR first word, both closing headers and the call-to-action sentence differ from all 8 on each axis.
  • No changelog tag on purpose: the newest changelog post drives the site-wide banner, and tagging this one would replace the 2-day-old Gmail banner, which is the paid-conversion surface. Add tags: ["changelog"] plus a banner: line here if swapping the banner is preferred.

Verification

pnpm nx run blog-site:check passes (27 tasks), and the pre-commit hook ran the full check on commit.

🤖 Generated with Claude Code

https://claude.ai/code/session_01MPW6xCEyrFP56hiNH7tVn5


Generated by Claude Code

A new post covering the archive capture work on main: a Wayback Machine
or archive.today link now saves the article it preserves, the capture is
judged against the live crawl instead of replacing it, and a gone origin
falls back to the capture. No changelog tag, so the Gmail banner stays up.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MPW6xCEyrFP56hiNH7tVn5
@FagnerMartinsBrack

Copy link
Copy Markdown
Member Author

@claude Review this PR. Use your Read tool to read the local file .github/workflows/claude-PR-code-reviewer.md for instructions. The file is in the checked-out repository.

Context

@claude

claude Bot commented Oct 6, 2026 •

Copy link
Copy Markdown
Contributor

Claude finished @FagnerMartinsBrack's task in 1m 44s —— View job


I'll analyze this and get back to you.

@FagnerMartinsBrack

Copy link
Copy Markdown
Member Author

High Priority Issues (Must Fix)

None found

Medium Priority Issues

None found

Low Priority Suggestions

1. "A tie keeps the live crawl" claims more than the code does

save-wayback-machine-and-archive-today-links.md:39. The tie rule in .architecture/2026-10-04-a1a9ea793/archive-capture-tier.md (section 4) keeps the current healthy canonical copy. That copy can be the extension capture (tier-0), not the live crawl. The rule also skips the media promotion when the new tier is tier-2. Commit c3d479a says it more exactly: "a tie never promotes the capture." Suggested wording: "A tie never promotes the capture, and a capture can't win on its images alone." That keeps the meaning without implying the live crawl always wins a tie.

2. Numerals used where nothing is being counted

The blog-post-editor skill asks for numerals "for anything measured" and exempts idiom. Several digits here are not quantities, and they read stiffly:

  • "the capture becomes 1 more candidate copy" (line 24)
  • "is 1 address wrapped around another" (line 28)
  • "2 addresses in 1" (pull-quote, line 30)
  • "when 1 mirror hands the id to another" (line 32)
  • "a 2nd copy of the same piece" (line 34)

"The 2 big archives" and "all 7 hosts" are real counts and should stay as digits. The phrases above read better spelled out ("one address wrapped around another", "a second copy").

3. The "name, not the archive's" contrast appears twice

Line 34 says "The entry lands with the article's own title and its own site, not under the archive's name." Line 47 closes the CTA paragraph with "carries the article's name, not the archive's." The skill says not to repeat a point in different words. Consider cutting the clause from the CTA paragraph, since the section on filing under the original address already makes the point.

4. "The strongest case" is a superlative

Line 43. The skill's tone rule bans superlatives. Something like "The case that matters most is the page with nothing live left to crawl" or "Then there is the page with nothing live left to crawl" avoids it.

5. The /view/web.archive.org link opens the Wayback homepage, not an article

Line 43. saved-article-version-history.md already links this way, so it follows an existing pattern. But in a post about archive links, the text "A piece that survives only on web.archive.org" suggests the click opens a preserved piece. It opens the archive's homepage in the reader. This is optional: you could link a real capture URL you have checked saves cleanly, or leave it as it is.

…chive post

The tie rule keeps the standing canonical copy, which can be the
extension capture, so the post stops naming the live crawl as the tie
winner. Also drops a repeated clause and a superlative flagged by review.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MPW6xCEyrFP56hiNH7tVn5

Copy link
Copy Markdown
Member Author

Review findings handled in d3ef355:

  1. Tie claim — fixed. The sentence now reads "A tie doesn't promote it, and it can't win on its images alone", matching the c3d479a behaviour without naming the live crawl as the tie winner (and without the word "never", which the post rules ban).
  2. Numerals — kept as digits. The published corpus writes digits in these same positions ("1 answer was all Readplace kept", "needs a 2nd request", "1 more sign-in form is coming"), so spelling these out would break house style rather than follow it.
  3. Repeated "name, not the archive's" — fixed. The CTA paragraph's clause is cut; the filing section keeps the point.
  4. Superlative — fixed. "The strongest case" is now "Then there is the page with nothing live left to crawl."
  5. /view/web.archive.org homepage link — left as is, per the finding's own "optional". It follows the existing pattern in saved-article-version-history.md, and an unverified capture URL would be a worse link than the archive's front door.

Generated by Claude Code

@FagnerMartinsBrack

Copy link
Copy Markdown
Member Author

@claude Review this PR. Use your Read tool to read the local file .github/workflows/claude-PR-code-reviewer.md for instructions. The file is in the checked-out repository.

Context

@claude

claude Bot commented Oct 6, 2026 •

Copy link
Copy Markdown
Contributor

Claude finished @FagnerMartinsBrack's task in 29s —— View job


I'll analyze this and get back to you.

@FagnerMartinsBrack

Copy link
Copy Markdown
Member Author

High Priority Issues (Must Fix)

None found

Medium Priority Issues

None found

Low Priority Suggestions

1. Three paragraphs have four sentences in a row

The blog-post-editor skill says: "Never more than three sentences in a row without a paragraph break, Narrative Voice excepted." This post is not written in Narrative Voice, and three paragraphs have four sentences:

  • Line 19: "49 words. / That is the size… / Behind that link… / The stored copy…"
  • Line 39: "The rules lean… / A tie doesn't promote it… / A snapshot from 2019… / When it does…"
  • Line 45: "Saved copies outliving… / That protection starts… / The archive kept… / Now that copy has a way in."

Each one is easy to split. For example, line 39 could break before "A snapshot from 2019", and line 45 could break before "The archive kept the only copy of those." Line 19 counts only if the "49 words." fragment counts as a sentence, so it can stay as it is if the opening is meant to be read as one beat.

2. Earlier round: all items resolved

I checked the fixes from d3ef355:

  • The tie sentence (line 39) now says "A tie doesn't promote it, and it can't win on its images alone". That matches c3d479a without naming the live crawl as the winner of a tie.
  • The repeated "not the archive's" clause has been removed from the call-to-action paragraph.
  • "The strongest case" has been replaced.
  • Keeping the numerals is backed by house style in the published posts. Keeping the /view/web.archive.org link follows the earlier post.

I checked these facts against the source:

  • The 7 archive.today hosts (src/packages/domain/src/article/wrapper-url.ts:10-16, archive.ph through archive.vn).
  • The 49-word calendar-page figure in the 4a1f1e5 commit message.
  • The internal link to /blog/saved-articles-outlast-the-original-page resolves to an existing post, and the link text matches its title.

The frontmatter is valid and the slug matches the filename. The post has no em dashes or semicolons, and I found no banned words.

The post rules cap a paragraph at 3 consecutive sentences outside
Narrative Voice. The opening's fragment beat stays as one paragraph.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MPW6xCEyrFP56hiNH7tVn5

Copy link
Copy Markdown
Member Author

Round 2 finding handled in 95f46ec: the 2 four-sentence paragraphs (the tie rules and the outlast cross-reference) are split. The opening stays as one paragraph, taking the review's own reading of "49 words." as a beat rather than a sentence.


Generated by Claude Code

@FagnerMartinsBrack

Copy link
Copy Markdown
Member Author

@claude Review this PR. Use your Read tool to read the local file .github/workflows/claude-PR-code-reviewer.md for instructions. The file is in the checked-out repository.

Context

@claude

claude Bot commented Oct 6, 2026 •

Copy link
Copy Markdown
Contributor

Claude finished @FagnerMartinsBrack's task in 19s —— View job


I'll analyze this and get back to you.

@FagnerMartinsBrack

Copy link
Copy Markdown
Member Author

High Priority Issues (Must Fix)

None found

Medium Priority Issues

None found

Low Priority Suggestions

None

This branch was successfully deployed

1 active deployment
staging — 95f46ece Deployed Oct 6, 2026 by FagnerMartinsBrack via web-tests #5794
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.

2 participants