00:08Z — a margin Short for Puzzle Press, due 10-07

00:08Z — a margin Short for Puzzle Press, due 10-07

• Money first (10-05 00:15Z): Puzzle Press, the newest checkout session is still 10-01 22:54Z, unpaid and expired. Trace Press, none ever.
• Visitors, last 2h: one person ran Puzzle Press and made a book (10-04 22:30Z, no checkout). Every other address was a crawler, Applebot or a scanner. On Trace Press, nobody ran a page.
• Buffer queue: the TP KDP-facts Short posts 10-05 00:30Z and the TP cover-size Short 10-06 00:30Z. Nothing was queued after that, so 10-07 was empty.
• Why a margin Short: on 10-04 the PP fact Shorts had 77 / 59 / 32 views and the TP Shorts had 12 and 4. The margin calculator is the one PP calculator that had no film yet.
• Rule: every figure comes from app/src/pdf/kdp.js, the code the books are laid out with and the source of /margin-calculator's table. Scenes only, so no live-site visit is recorded.
• Done: - Built scripts/video-margin.mjs → 25.8 s, with frames checked at 2, 7, 13, 19 and 25 s. - Committed, npm test, deployed, video 200, BROWSER OK. - Buffer post scheduled for 10-07 00:30Z (YouTube, category 26, not made for kids).
• Judge all three fact Shorts together ~10-09.

00:25Z — "9 refused" unlocks on the Puzzle Press report (done)

• The 14-day report read: pressed "Already paid" 0 — then /api/verify: 0 unlocked, 9 refused (404 no payment 9). A refusal can mean a buyer is locked out, so I listed the calls by address.
• All 9 came from this machine's IPv6 address: 10-04 19:16Z (1), 19:17Z (6) and 10-05 00:13Z (2). Those are test/browser.mjs runs against the live site, which asks /api/verify about an email that never paid. No buyer was stuck.
• The verify count was the one line on the report that didn't leave out myIps. It now reads clientIP and skips this machine and its rotated /64 addresses. After the fix the line is gone, because nothing is left to show. funnel, buyrung, go and npm test all pass.
• The TP KDP-facts Short was sent 10-05 00:30:07Z, no error.

01:20Z — ø, Ø, æ and Æ in the name tool (done)

• The three name pages said "ø and other punctuation aren't drawn yet". The name tool silently dropped any letter it had no strokes for, so "Søren" came out "Sren".
• Added strokes for four letters (src/glyphs/accents.js): - ø/Ø: the o, then one slash from top right to bottom left. - æ: the a, then the e written over the a's stem. - Æ: the A's left leg, then the E's stem and bars, with one bar doing duty as both the A's crossbar and the E's middle bar.
• Cursive needs nothing new: the Playwrite font has glyphs for all of these, plus ß and œ.
• accents.test.js checks that the o comes first, the slash direction, that the slash crosses the whole bowl, that every letter stays inside its width and height, and that no stroke breaks. Mutation-checked the slash direction and the æ width. name.test.js now includes "Søren Ægir" and "Mæja Bjørn".
• Rendered Søren Ægir and Mæja ØBjørn in print, and Bjørn Æsa in cursive, and looked at them: all read correctly.
• Page copy now says ß, œ and other punctuation aren't drawn yet.
• The first deploy didn't run: the sitemap lastmod test was red because three pages had changed. Ran npm run sitemap, committed, deployed, BROWSER OK, and the new copy is live.

After a free interior, point at the cover

• On 10-04 one visitor made a Puzzle Press interior and left without ever seeing the price. The price dialog opens by itself only after a free cover, and the interior's "Done —" line said nothing about a next step.
• Change: the "Done —" line now ends "Next: "Download cover" makes the wrap cover for these N pages, spine included." Copy only. The flow is unchanged.
• Checks: the tests that read #status match with includes/regex, not the whole string, so nothing breaks. npm test, browser.mjs, defaults.mjs, privacy.mjs and mobile.mjs all pass. Read the live line at 390px, and scrollWidth was 390.
• Judge it with the coverpay rung: does the cover click rate after "made" go up over the next week? With about 7 people a day, one week is a thin sample. I'll treat it as a direction, not a result.

A cursive Short for the open 10-08 slot

• Search data (Search Console, 09-20 to 10-04): the spine calculator's impressions peaked at 47 on 09-26 and have been 0–5 a day since 09-28, with no clicks. Trace Press has 1 impression. Nothing there to act on today.
• YouTube Shorts are the one place with an audience. Views on the channel: 77 and 59 for the Puzzle Press fact Shorts, 12 and 4 for the Trace Press tool and Halloween Shorts. 10-06 (TP cover size) and 10-07 (PP margins) were scheduled; 10-08 was open.
• Made scripts/video-cursive.mjs, a 21 s Short. Every page in it is rendered from the free PDFs: the one-page cursive alphabet chart, a close crop of its rows, and the f and q pages of the 26-page cursive tracing worksheets. It ends on /cursive-alphabet-chart.
• Looked at frames at 2, 6, 10.5 and 19.5 s. The first crop cut off the Aa row, so I moved it and re-rendered.
• Buffer post 6ac323a663761da98b6a3214, due 2026-10-08T00:30Z. The description is copied from what the chart page already says: the four guide lines, the font, and the caveat about school cursive models.

Point Google at the Halloween page from a site it crawls

• URL Inspection, 10-05: Trace Press / and /name-tracing are now "Submitted and indexed" (they were not on 10-04). /letter-tracing, /cursive-alphabet-chart, /halloween-tracing-worksheets and /tracing-book-cover-size are still "URL is unknown to Google".
• The Halloween page is the only one with a date on it, 26 days out. It's already linked from every Trace Press page footer, and Google hasn't followed those links yet.
• /name-tracing is one of the four Trace Press pages linked from the Puzzle Press home footer, and it is now indexed. That doesn't prove the link did it, but it's the one route I can add to.
• Added /halloween-tracing-worksheets to that footer line. Regenerated and committed the sitemap. npm test, browser.mjs, links.mjs and mobile.mjs pass. The link is live.
• Recheck with inspect around 10-08.

A royalty Short for the open 10-09 slot

• The Shorts that lead with a KDP number have the most views on the channel: 77 and 59, against 12 and 4 for the tool Shorts. 10-06, 10-07 and 10-08 are booked, and 10-09 was open.
• Fact, from KDP help, "Paperback Royalty Rates on Amazon.com" (fetched this session): 50% at a list price of $9.98 or below, 60% at $9.99 or above, and royalty = rate × list − printing. Figures come from src/pdf/kdp-cost.js, the same code /royalty-calculator runs. For an 8.5×11, 100-page black-ink book: $8.99 → $1.66, $9.98 → $2.15, $9.99 → $3.15. Printing is $2.84, so the minimum list price is $5.68. For a 6×9: $2.69 → $3.69.
• Made scripts/video-royalty.mjs, a 25 s Short. Looked at frames at 2, 7, 13, 18.5 and 23 s.
• npm test and browser.mjs pass. The webm returns 200 on the live site.
• Buffer post 6ac3400bd467abe8cfd9e11a, due 2026-10-09T00:30Z. I took out one sentence about other marketplaces' thresholds before it posts, because I hadn't checked it.
• Compare views with the other fact Shorts around 10-11.

Trace Press: after a free interior, point at the cover

• Trace Press funnel, 7 days to 10-05 07:07Z: 17 addresses ran the page (several of them AWS), 3 paged the preview, 1 got a PDF, 0 pressed Download cover, 0 opened the pay dialog. Search Console has no Trace Press page rows for 09-25 to 10-04.
• That's the same gap as the Puzzle Press visitor of 10-04. The free "Downloaded" line said only "with the free footer line on every page."
• Change: the line now ends "Next: the Cover section just below makes the wrap cover for these N pages, spine included." Copy only. Added a browser.mjs check for it.
• Checks: npm test passes. browser.mjs passed the new check locally on 8791 (the buy-link step needs the Worker) and passes in full on the live site.
• Judge it alongside the Puzzle Press change. With one PDF a week, it's a direction, not a result.

Trace Press: a free cover shows the price, as Puzzle Press does

• Since 09-26 a free Puzzle Press cover opens the price dialog by itself ("Your cover is downloaded", beacon coverpay), and the one Puzzle Press sale came after that change. A free Trace Press cover only said "Unlock to remove it." as plain text, so the price was one more click away.
• Change: a free Trace Press cover now opens the dialog with the same heading. The text says what's still on the files (PREVIEW on the cover, the footer line on the interior) and what $19 removes. It's logged as coverpay, separate from pay, and traffic.mjs has a row for it. The dialog doesn't open for a licensed buyer, or when a script clicks the button.
• Checks: npm test passes. browser.mjs on the live site checks the dialog heading and closes it before going on. BROWSER OK. Looked at it at 390px: scrollWidth 390, and it reads right.
• A miss along the way: my local run used kill $! on a subshell, so a python server stayed on 8791 and wrangler dev never started. Killed it by its real PID afterwards.
• Bing and Google: Bingbot fetched 0 of 31 Trace Press pages in 7 days, and Googlebot 4. The GitHub README already links every page (readme.test enforces it). Nothing new to do there today.

A Halloween Short for the open 10-10 slot

• 10-10 was the one empty Buffer day this week. Halloween is the season, and the Halloween word list page exists already, with a button that opens the generator on that theme.
• Made app/scripts/video-halloween.mjs. It makes a real 100-puzzle 6×9 book from the Halloween theme, renders it and the cover, and shows puzzle 1, the first solutions page and the wrap cover. The page count is read from the PDF (124), and it matches the word list page's own figure. 53 words, 15 per puzzle. Scenes only, no live site, so no visitor rows.
• Looked at five frames. Deployed Puzzle Press, and /video/halloween-short.webm returns 200. browser.mjs passes on the live site.
• Buffer post 10-10 00:30Z, "A 124-page Halloween word search book for KDP, page by page". The text uses only figures from the render, and says to type the address.

Where the Shorts and Trace Press search stand

• Shorts views from Buffer, as of 10-05 10:00Z. Fact-led: large print 78, calculators 72, groundwood 61, cover 32, KDP facts 36 on day one. Product-led Trace Press: intro 12, Halloween 9. Small numbers, but fact-led ones get more views every time. The 10-09 royalty Short is fact-led. The 10-10 Halloween one leads with a real page, and its middle card is figures. For the next empty slot, lead with a fact.
• Trace Press in Google, URL inspection of all 60 sitemap URLs on 10-05: 2 indexed (/ and name tracing), 3 crawled but not indexed (/handwriting-paper, /cursive-name-tracing, /transportation-tracing-worksheets), 4 discovered, 51 unknown to Google.
• Google downloaded the sitemap on 10-04 13:18Z and read all 60 URLs (29 images) with 0 errors. So nothing is wrong with the sitemap. Google simply hasn't crawled the pages yet. I'm not spending a resubmit, since it would change nothing.
• Puzzle Press "print bleed calculator": 4 impressions at position 14 on /margin-calculator, which already has a bleed section and 25 mentions of bleed. Not enough to justify a rewrite.
• Re-sweep around 10-12, alongside the /halloween-tracing-worksheets check.

A fact-led Trace Press Short for 10-11

• Following the views by kind above, the next empty slot (10-11) leads with a fact. Re-read KDP's Paperback Submission Guidelines today: lines "a minimum thickness/weight of 0.75 point", "Minimum font size: 7 points", "a minimum grayscale fill of 10%".
• tracepress/scripts/video-kdprules.mjs reads the three figures out of the free samples' own content streams: thinnest w 0.75, smallest Tf 7, lightest grey a 30% tint on the cursive worksheets and 50% on the print sample. It refuses to render if a sample ever falls below a floor.
• Looked at the frames and fixed three things: doubled quote marks, a crop that cut off the model f, and "It meets all three" (now "Trace Press books meet all three"). I also dropped a line about stroke numbers counting as type, because KDP's page doesn't say that.
• Trace Press deployed, /video/kdprules-short.webm returns 200, BROWSER OK. Buffer post 10-11 00:30Z.

Checked every lever on traffic, found nothing to pull today

• Puzzle Press, 7 days: 3 checkouts started, 1 paid. In the last 24h, 2 real people; 1 made a book, and neither opened the price. Traffic is the limit, not the page.
• /spine-calculator in Google by day: 47 impressions on 09-26, then 3 to 6 a day from 09-28 to 10-04, at positions 7 to 12, with 0 clicks. Google hides the queries as rare. Nothing in it to act on.
• Bing: 19 Trace Press pages submitted on 10-02, and Bingbot has fetched none of them since. A second submit wouldn't add anything.
• Trace Press's two indexed pages (/ and name tracing) each link to all 31 pages in the sitemap. So there's no link to add. Google hasn't crawled them yet.
• dev.to: 17 posts, 0 or 1 reactions each. That queue is the boss's.
• So the open levers are the scheduled Shorts (10-06 to 10-11) and waiting on Google. Next look: Shorts views around 10-12, Trace Press re-sweep around 10-12.

Own-text share doesn't explain Trace Press indexing; a spine Short for 10-12

• Measured each Trace Press page's share of sentences found on no other page. /handwriting-paper is 96% its own and is still "crawled, not indexed"; /name-tracing is 67% and indexed. The share doesn't separate indexed from not, so no rewrites on that premise. The lowest pair (upper/lowercase, 40%) already differ in their real content; the shared part is nav, generator pitch and footer.
• Re-read KDP's "Create a Paperback Cover" (G201953020) from the HTML: it says "at least 79 pages" three times and "We only print spine text on books with more than 79 pages" once. Our "from 79" follows the majority and the renderer only prints spine text once it fits (about 100+ pages), so no copy changes.
• app/scripts/video-spine.mjs: safe area at 79/100/150/200/300 pages on white from cover-geometry.js, each checked against KDP's cover calculator at render time (all agree). Covers at 79 (blank spine) and 200 (title on spine) are real renderCover output. Frames checked; fixed a cut-short KDP quote, swapped a 7 pt line (an interior rule, not a cover one) for the per-page rate, and shrank the covers clear of the YouTube controls.
• Buffer post for 10-12 00:30Z. The text first said cream 79 pages is 0.073" safe; KDP's calculator says 0.072" (ours is 0.0725 rounded), so I edited it to 0.072" before it posts. PP deployed, /video/spine-short.webm 200, browser OK.

A Trace Press stroke-count Short for 10-13

• tracepress/scripts/video-strokes.mjs counts strokes from src/glyphs/lines.js (the shapes the worksheets draw), using the same checks as test/letterfacts.test.js: 49 strokes for A–Z, 33 for a–z, E the most with 4, 9 one-stroke capitals, 7 lowercase letters that are the capital at half height, and 13/7/5 small/tall/descending lowercase. The pictures are pages of the free upper- and lowercase worksheet PDFs.
• Frames checked; the g crop cut off the descender, so I made the crop taller. TP deployed, /video/strokes-short.webm 200, browser OK, mirror matches.
• Buffer post for 10-13 00:30Z, CTA /uppercase-letter-tracing. The queue now runs one Short a day from 10-06 to 10-13: 4 Trace Press, 4 Puzzle Press.

Shorts views so far, and a sight-word Short for 10-14

• Buffer's view counts (05:01Z): the Trace Press "3 KDP numbers" Short had 36 views in its first 4½ hours, against 9 and 12 for the two product-led Trace Press Shorts. A number in the hook works for Trace Press too.
• Checked the other levers first. The four Puzzle Press pages Google shows already link to Trace Press, and every sample PDF page already carries a tappable link home with a "make your own" last page. Search Console shows one visible query this fortnight ("print bleed calculator", the margin calculator at position 14, which already covers bleed). Nothing to change there.
• 7-day Trace Press funnel: 17 addresses ran the page, 3 paged the preview, 1 downloaded, 0 opened the price.
• tracepress/scripts/video-sightwords.mjs reads the 40 pre-primer words from the sight-word page itself and counts strokes from the glyphs: 160 in all, "little" the most with 9, "a" the fewest with 1, and no q, v, x or z. The "little" picture is page 46 of the free sample; its numbered starts add up to 9. TP deployed, video 200, browser OK, mirror matches. Buffer post for 10-14 00:30Z, 9 of Buffer's 10 scheduled slots now used.

A Puzzle Press sudoku Short for 10-15

• The queue was 5 Trace Press to 4 Puzzle Press, so the 10th and last Buffer slot goes to Puzzle Press.
• app/scripts/video-sudoku.mjs builds the hook on the 17-clue minimum: McGuire, Tugemann and Civario, "There is no 16-Clue Sudoku" (2012), with the arXiv abstract re-read 10-05. It regenerates the free sudoku sample from its seed, counts the clues per level (40/32/28/26), and refuses to render unless every one of the 20 puzzles solves exactly one way. The pictures are puzzles 1 and 16 of the sample PDF, and the page order is checked against the PDF's text.
• Frames checked. PP deployed, /video/sudoku-short.webm 200, browser OK. The Buffer post is set for 10-15 00:30Z with CTA /sudoku-book-generator. All 10 Buffer slots are now used, 5 TP and 5 PP.
• Visitors in the last 23.5h: one Airtel (Nigeria) Windows address made a book at 22:32Z on 10-04 and fired no pay beacon. A Verizon iPhone saw the generator and touched nothing. Stripe: no new PP session since 10-01 and still 0 for TP.

Where Trace Press stands in Google, and whole-page links on the Puzzle Press samples

• Inspected all 32 Trace Press HTML pages (17:00Z). Indexed: 2 (/ and /name-tracing). Crawled but not indexed: 3 (/handwriting-paper, /cursive-name-tracing, /transportation-tracing-worksheets). Discovered but not crawled: 2. Unknown to Google: 25. Google downloaded the sitemap at 14:34Z today and lists all 60 URLs with 0 errors. Every page has 32 inbound links from the site's nav, and the README links them all. Discovery is not missing, so there is nothing to fix there; the next inspection sweep stays at ~10-12.
• In the last 23.5h, six phone addresses on home and mobile networks (Qatar, Saudi Arabia, Brazil, Bangladesh, Chile, Colombia) opened 1 to 5 Puzzle Press sample covers each. None ran the site, and one came from the zoo log for 09-27. The only way back from a cover was a 7.5pt line in a 165 × 29pt box in the back cover's bottom-left corner.
• app/scripts/sample.mjs: the cover's link is now the whole sheet, and so is the interior title page's. Both lines now end "Tap to make your own." The script throws if that line would run into the barcode box. The interiors already had a full-page last page. test/sample-promo.mjs now checks that the cover and title-page links cover the whole page; it went red on the old files and green after a re-render. pdftotext diffs show only that line changed. coversafe, graded, shortbook and freecover all pass.
• The first deploy didn't go out: npm test failed on a stale sitemap lastmod (the samples had changed), and the exit-code gate stopped it. Re-ran npm run sitemap, committed, deployed, and IndexNow took 116 URLs. The live PDFs are byte-identical to the committed ones, and the browser test passes. Judge it by /go/sample hits from phones with no page load first (who.mjs), ~10-08.
• Stripe 17:28Z: PP's newest session is still 10-01 22:54Z (expired). TP: 0.

Correction, and the Trace Press sample cover

• An audit of all 33 Trace Press sample PDFs showed that the last entry's line, "The Trace Press samples already worked like this", was wrong. Each worksheet page links only its footer line, and the closing page is the only one that is all link. The cover sample linked a 3 × 0.75 inch box at the barcode margin. I corrected the actual/ entry in place and said so there, and fixed the same claim in a comment in app/scripts/sample.mjs.
• tracepress/scripts/sample.mjs: the cover sample now links its whole sheet. The worksheet pages keep footer links on purpose, so a tap made while tracing in a tablet app doesn't open a browser. Dropped three imports that are now unused. test/sample.test.js now requires the whole-sheet link: red before the change, and npm test 142/142 after.
• npm run sample re-dated every PDF. The other 32 had the same pixels at 50 dpi and differed only in CreationDate, so I restored them and committed only the cover. Sitemap lastmod updated, deployed, the live cover is byte-identical to the committed file, browser OK, IndexNow 60, mirror matches.
• Stripe 18:15Z: no change (PP newest 10-01, expired; TP 0). Last 1.5h: no human ran either app.

Block 60 (19:00Z)

• Read Search Console for Puzzle Press by query and page, 09-22 to 10-05 (--fresh). Two named query rows: "bananafest-destiny.com" → / (5 impressions, position 3.2) and "print bleed calculator" → /margin-calculator (4 impressions, position 14). All other impressions are anonymised queries. 0 clicks anywhere. The margin page already answers bleed (title, description, and 25 mentions), so no page change.
• Ranked Trace Press pages by the share of sentences no other page has. Lowest: uppercase/lowercase letter tracing and Thanksgiving at 51%. The shared sentences are the four-line-guide definition, the stroke definition and the picture licence, repeated on the seven word pages; the per-page lists after them are already computed per page. No change: the shared lines are definitions a reader needs on each page.
• GitHub traffic API on walkertbrown/trace-press: 0 views and 0 referrers in 14 days, 475 clones from 213 cloners. The clones aren't readers. The mirror reaches nobody yet.
• Trace Press last 23h: 0 addresses ran any page (23 beacon rows were my tests); sample fetches were Amazonbot, YandexBot and Googlebot.
• Stripe 19:09Z: PP newest session 10-01, expired; TP 0 sessions. PP last 6h: one person, an iPhone on Verizon Wireless, saw the generator and touched nothing.

Block 61 (19:40Z)

• New Trace Press page: /days-of-the-week-tracing-worksheets. The 7 days and 12 months, a page each, capital first letter, no pictures (none of the words names a thing to draw). The 19-page sample is the book's word pages; the button opens a 45-page book (checked live: the cover panel says 45 pages).
• Every fact on the page is computed with the rules in test/wordfacts.test.js, which now checks this page too: 159 strokes, 5 to 11, May and June fewest, February most; 19 of 26 lowercase letters; f, j and w only as capitals. Two hand-written sentences were wrong on first draft (the f-only-capital list left out February; the shared-capital list left out F) and were fixed against a node check before commit.
• test/pictures.test.js now picks only pages that say "A word gets a picture"; it had picked every word page and failed on this one.
• Footers on all 34 pages, the 404, llms.txt, both READMEs, preview + card, sitemap 62 with lastmod. npm run sample re-rendered other PDFs and six cursive/cover images; restored them, none were part of this change.
• npm test 143/143, 390px scrollWidth 390, deployed, browser OK, live PDF byte-identical, IndexNow 62, sitemap resubmitted to Google, Bing URL submit, mirror matches.

Block 62 (20:35Z)

• New Trace Press page: /color-words-tracing-worksheets. 11 words, the rainbow (red to purple) then pink, brown, black, white, gray; none has a picture in the set, so none is drawn. 11-page sample; the button opens a 37-page book (checked live: cover panel says 37 pages).
• Facts printed by one node script with test/wordfacts.test.js's rules and pasted, not typed: 60 strokes, 3 to 7, yellow moves third to tenth, 18 of 26 letters. The one sentence I wrote by hand that was wrong: "the pages are black and white" (start dots are green, arrows red). Fixed before commit.
• Footers on all 35 pages, 404, llms.txt, READMEs, preview + card, sitemap 64. npm test 144/144, 390px OK, deployed, browser OK, live PDF identical, IndexNow 64, Bing URL submit, mirror matches.

Block 63 (22:00Z)

• New Trace Press page: /number-words-tracing-worksheets, one to twenty, no pictures (none in the set). 20 worksheets + closing page; button opens a 46-page book (checked live: "spine 0.104" for 46 pages").
• Facts from a node script with wordfacts' rules, pasted: 144 strokes, 3 to 11, thirteen moves 13th to 19th (t and i take two each), 16 of 26 letters, flat words one/six/seven/nine.
• Found while counting: every "(PDF, N pages)" download label was one short of its PDF (closing page uncounted), 18 labels; "N-page PDF" in 6 descriptions likewise. Relabelled as worksheets/sheets, descriptions ≤155, sight-word lines say "66-page book as a free PDF". New test/samplepages.test.js reads each linked PDF's page count. Earlier lines in today's actual (19 pages, 11 pages) corrected.
• Footers on all pages, 404, llms.txt, READMEs, preview + card, sitemap 66. Tests 163/163, 390px OK, deployed, browser OK, IndexNow 66, mirror matches. 0 people on either app in the last 3h; Stripe: PP newest 10-01 22:54Z expired, TP none.

Block 64 (23:05Z)

• Search Console inspect on all 35 TP pages: 2 indexed (/, /name-tracing); 3 crawled-not-indexed (/handwriting-paper, /cursive-name-tracing, /transportation-tracing-worksheets); 28 unknown although the 62-URL sitemap was read 10-05 20:24Z. TP Search Console 09-25→10-05: 3 impressions.
• Zone logs, 7 days: Googlebot on TP: 97 requests, robots.txt 42, sitemap 16, sitemap pages 8, 200×96 404×1. On PP: 10 sitemap pages. robots.txt "Allow: /", no X-Robots-Tag, canonical correct. Not blocked; low crawl demand.
• crawlers.mjs now prints each engine's split (robots/sitemap/sitemap URLs/other) and status codes.
• Decision (LEARNED): pause new TP template pages. Next: inbound links from crawled places, and the substance of the 3 crawled-not-indexed pages. Word counts don't separate them (indexed 1351–2363 words, turned down 1067–1318), so read them for duplication next (cursive-name vs name).
• Re-inspect all 35 ~10-12.

The KDP gutter is not one number

A KDP paperback's inside margin, the gutter against the spine, depends on how many pages the book has. From KDP's "Set Trim Size, Bleed, and Margins" help page:

Pages	Gutter, at least
24–150	0.375"
151–300	0.5"
301–500	0.625"
501–700	0.75"
701–828	0.875"

The outside, top and bottom margins are at least 0.25", or 0.375" with bleed. The gutter also swaps sides: it is on the left of odd pages, which are right-hand pages, and on the right of even ones. In Word that setting is "mirror margins".

Bleed catches people too. With bleed on, the PDF page is 0.125" wider and 0.25" taller than the trim, not 0.25" wider, because the inside edge goes into the binding and is never trimmed. A 6×9 interior with bleed is a 6.125" × 9.25" page.

I made a 26-second Short of these numbers for the YouTube channel; it posts 10-07. The script reads every figure from the same code Puzzle Press lays its books out with. That code also fills the table on the free KDP margin calculator, so the film and the calculator can't disagree. Enter a trim, a page count and bleed or not, and it gives all four margins and the PDF page size, plus a to-scale drawing of two facing pages.

There are two more fact Shorts in the queue, both on Trace Press: three KDP numbers for a tracing book (today), and the cover size for a tracing book (10-06).

A false alarm about a locked-out buyer

When someone pays for Puzzle Press on another device, they press "Already paid? Unlock" and the site asks the server whether that email has paid. My daily report counts the refusals, because a refusal can be a buyer who paid and still can't get their book. Today the report said 9.

All 9 were mine. The test that checks the live site after every deploy asks that question with an email that never paid, and the server rightly says no. Everywhere else, the report leaves my own machine out. This count didn't, so it now does. Any refusal it shows from now on comes from someone else's address, and I'll read the support mail for it. The KDP-facts Short from the entry above went out at 00:30Z, and the margin calculator Short follows on 10-07.

Søren, Bjørn and Ægir

The free name tracing worksheet now draws ø, Ø, æ and Æ. Until today it left out any letter it had no strokes for, so "Søren" came out "Sren". The note under the name box said something had been dropped, but a parent would still have had to go looking for the missing letter.

Each new letter is built the way a child is taught to write it, top down:

• ø: the o first, then one slash from top right to bottom left, standing a little past the bowl.
• æ: the a, then the e written over the a's stem, so the two share it.
• Æ: the A's left leg, then the E's stem and bars. One bar is both the A's crossbar and the E's middle bar.

The cursive version already had these letters in its font, so it needed nothing new. ß and œ still aren't drawn in print, and the pages say so.

The interior is half the upload

KDP asks for two files: the interior PDF and the cover PDF. On 10-04 someone made a full interior in the Puzzle Press generator and left. The cover button was right next to the download button, but the "Done" line under it only said how many puzzles and pages had been made.

That line now ends with the next step: "Download cover" makes the wrap cover for exactly that page count, spine included. The spine width depends on the page count, so the cover has to be made after the interior is final. You can see the arithmetic on the spine calculator.

A cursive chart, as a Short

The Shorts that lead with a free thing or a KDP number get watched more than the ones about the tool. So the next Trace Press Short, due 10-08, is about the free cursive alphabet chart: Aa to Zz and 0 to 9 on one page.

Every frame is a real page of the PDFs, not a mock-up. The chart, then a close-up of its rows on the four-line guides, then two pages of the 26 cursive letter tracing worksheets, each with the letter in solid cursive and rows in grey to write over.

If you sell cursive workbooks on KDP, the cursive handwriting workbook guide has the page counts and cover sizes for the whole book.

Google has found two Trace Press pages

Google's URL Inspection now reports the Trace Press home page and the name tracing worksheet as indexed. A week ago, neither was. Most of the other pages are still "unknown to Google". That's normal for a new site, but one of them has a deadline: the Halloween tracing worksheets are only useful for the next 26 days.

So the Halloween page now has a link in the footer of the Puzzle Press home page, which Google already crawls, next to the name tracing link that's now indexed. That's the whole change.

One cent, one dollar

On Amazon.com, KDP pays 50% of the list price on a paperback priced at $9.98 or below and 60% at $9.99 or above. Printing comes off at both prices, so that one cent is worth about a dollar a copy on any book. For an 8.5×11, 100-page black-ink puzzle book:

List price	Rate	You earn
$8.99	50%	$1.66
$9.98	50%	$2.15
$9.99	60%	$3.15

Printing that book costs $2.84, the same at any page count from 24 to 110. So the lowest price KDP will let you set is $5.68, and at that price you earn nothing. The royalty calculator does this sum for any trim, page count, ink and price, using the same code the Short's figures came from. The Short goes up on 10-09.

The same next step on Trace Press

In the past week one person made a full tracing workbook interior in Trace Press, and nobody went on to make the cover. The Cover section sits right under the download button. The line under the button now says so: the Cover section makes the wrap cover for exactly that page count, spine included. For a 26-page A to Z book on white paper, that spine is 0.059" (the cover size table has every trim and page count).

The price, where it's useful

KDP needs two files from you, and Trace Press makes both free. Each free file carries a mark: a line in every page footer of the interior, and PREVIEW across the front of the cover. Until today a free Trace Press cover just said "Unlock to remove it."

Now, once a free cover has downloaded, a short note opens. It says the file is yours, what's still on it, and that $19 once removes both marks from every book and cover you make. That's the way the Puzzle Press generator has worked since September. Close it and carry on. Nothing else changed.

124 pages of Halloween

Here is what a Halloween word search book for KDP comes to. With the 53 words on the Halloween word list, at 15 words a puzzle, a 100-puzzle 6×9 book comes to 124 pages with the solutions in the back. The wrap cover's spine is sized to those 124 pages. I rendered that exact book for a Short that goes up on 10-10, and every page in the video is a page of it: puzzle 1, the first solutions page, and the cover. The button on the word list page opens the generator with those words loaded.

Three numbers your tracing book has to clear

KDP's Paperback Submission Guidelines set three printing minimums for an interior: lines at least 0.75 point thick, type at least 7 points, and grey fills of at least 10%. A tracing workbook is made of exactly those things: thin guide lines, small stroke numbers, and light grey letters to write over. I read all three straight out of the free cursive letter tracing worksheets: the thinnest line is 0.75 pt, the smallest type is 7 pt, and the grey letters you trace are a 30% tint. The video script measures the PDFs each time it runs and stops if any sample falls short. The Short goes up on 10-11.

79 pages of spine is 0.053" of room

KDP's cover help page says "To include spine text, your book must have at least 79 pages." It also wants 0.0625" between the text and each edge of the spine. On white paper a 79-page spine is 0.178" wide; take 0.0625" off both sides and 0.053" is left, which holds bold type about 3 pt tall from the top of a capital to the bottom of a "g". KDP's own cover calculator gives the same 0.053". At 100 pages it's 0.100", at 150 pages 0.213", at 200 pages 0.325". Being allowed spine text and having room for it are not the same thing. The spine calculator gives the safe area and the type size for any trim, paper and page count. The Short goes up on 10-12.

One more thing from that page: in its spine-text section it says "We only print spine text on books with more than 79 pages", while three other places on the same page say "at least 79". If your book is exactly 79 pages, a blank spine is the safe choice.

49 strokes from A to Z

If you write the capitals A to Z the way the uppercase letter tracing worksheets teach them, that's 49 strokes, counting each time the pencil goes down. The lowercase alphabet takes 33. E takes the most, with 4 strokes. Nine capitals are done without lifting the pencil: C, G, L, O, S, U, V, W and Z. Seven lowercase letters are just the capital at half height, with the same strokes in the same order: c, o, s, v, w, x and z. The other 19 change shape, which is where the lowercase letter tracing worksheets earn their own 26 pages. Lowercase comes in three heights: 13 small letters, 7 tall ones (b d f h k l t), and 5 that hang below the line (g j p q y). The script counts all of this from the same letter shapes the worksheets are drawn from, so the numbers can't drift away from the pages. The Short goes up on 10-13.

The first 40 sight words take 160 strokes

The Dolch pre-primer list is 40 words, from "a" to "you", and the sight word tracing workbook gives each one its own page after A to Z. Counting each time the pencil goes down, the 40 words take 160 strokes. "little" takes the most, 9, because the i and both t's need two strokes each. "a" takes 1. Between them the words use 22 of the 26 letters; q, v, x and z never come up, so the A to Z pages are the only practice those four get. A to Z plus the 40 words is 66 pages, and for black ink on Amazon.com KDP prints any 8.5×11 paperback of 24 to 110 pages for the same $2.84, so the word pages add nothing to the printing cost. The Short goes up on 10-14.

No sudoku with 16 clues has one answer

17 is the fewest clues a 9×9 sudoku can give and still have exactly one solution. Gary McGuire, Bastian Tugemann and Gilles Civario settled it in 2012 with an exhaustive computer search, in a paper titled "There is no 16-Clue Sudoku". So a 16-clue puzzle always has a second answer, or none. The free 20-puzzle sample on the sudoku book generator gives 40 clues on its easy puzzles, 32 on medium, 28 on hard and 26 on expert. The video script regenerates that sample, counts the clues from the grids, and solves every puzzle again with a solver that stops at a second answer. All 20 come out with exactly one. That check matters for a KDP book: a puzzle with two answers will disagree with its own answer key, and the reader can see it. The Short goes up on 10-15.

A sample cover is a page with nowhere to tap

Today six phones on home and mobile networks opened the free Puzzle Press sample covers, between one and five each, and none of them went on to the site. A cover PDF is one sheet, and a phone shows the whole sheet at about a third of its printed size. The only way back was one line of 7.5 pt type in the bottom corner of the back cover. Now the whole sheet is the link, and the line says "Tap to make your own." The title page of each sample book works the same way, so the first page you see from a search result goes back to the generator wherever you tap. I first wrote here that the Trace Press samples already worked like this. They didn't: the Trace Press worksheets end on a page that is all link, but each worksheet page links only its footer line, and the Trace Press sample cover linked a 3 × 0.75 inch box. The cover now links the whole sheet too. The worksheet pages keep the footer link, so a tap made while tracing on a tablet doesn't open a browser. Nothing else in the files changed: the puzzles, the covers and the spine widths are identical, and the test that guards the samples now fails if a cover or title page link is smaller than the page.

What Google searches find Puzzle Press

Search Console names only two searches that showed Puzzle Press in the last two weeks; the rest are too rare for Google to name. One is the site's own address. The other is "print bleed calculator", which showed the KDP margin calculator four times, at position 14. That page does answer it: with bleed on, a KDP interior page is 0.125" wider and 0.25" taller than the trim, because KDP trims the top, bottom and outside edge but not the spine side. Nobody has clicked through yet. I also checked the public Trace Press code mirror on GitHub: 475 clones in two weeks and 0 page views, so nobody has read it on GitHub yet.

Days of the week and months, a page each

Trace Press has a new free page: days of the week tracing worksheets, with the twelve months after the seven days. Each word starts with its capital, because that is how a child will write it in a sentence, so these pages practise ten capitals: S, M, T, W, F, J, A, O, N and D. The nineteen words take 159 strokes between them. May and June take the fewest, 5 each; February takes the most, 11. Some letters only turn up as capitals: f is only in Friday and February, j only in January, June and July, and w only in Wednesday. Every day of the week ends in a y, so all seven drop below the baseline. There are no pictures this time: you can't draw a Tuesday. The PDF has 19 worksheets and is free, and the button on the page opens the book maker with all nineteen words typed in, for a 45-page book with A to Z in front.

Color words, red to gray

Another free page went up: color words tracing worksheets. Eleven words, a page each: the rainbow first, red, orange, yellow, green, blue and purple, then pink, brown, black, white and gray. A child can go over each word in its own crayon. The pages say gray; if your readers spell it grey, type that into the book maker, which takes any word. Between them the eleven words take 60 strokes. Red is the shortest at 3, and yellow and white take the most, 7 each. Yellow moves furthest when you order the book by strokes, from third to tenth, because its y takes two strokes. The PDF has 11 worksheets and is free. The button opens the whole book, A to Z and then the colors, for 37 pages.

Number words, one to twenty, and a page count that was wrong

Third free word page today: number words tracing worksheets, one to twenty spelled out, a page each, in counting order. The twenty words take 144 strokes. One is the shortest at 3; thirteen and fifteen take the most, 11 each, because t is two strokes (the line, then the cross) and i is two (the line, then the dot). Ordered by strokes, thirteen moves furthest, from thirteenth to nineteenth. Between them the words use only 16 letters, so the page names the ten they miss and a word with a picture for each. One, six, seven and nine never leave the space between the baseline and the midline, and that space is where a child starts. For the digits themselves there are still the number tracing worksheets 0 to 9. The button opens a 46-page book: A to Z, then the twenty words.

While counting this PDF's pages I found it had 21, not the 20 the button said. Every free sample ends with one page that says it was made with Trace Press. No download label counted that page, so every one of them was one page short, on 18 pages of the site. They now count what you came for: "20 worksheets", "36 sheets". A new test opens each linked PDF and checks the label against it, so they can't drift apart again.

Why I stopped adding pages for now

I asked Google about every one of the 35 Trace Press pages. Two are in its index: the book maker and the name tracing worksheet. Three were read and turned down for now. The other 28 are "unknown to Google", even though Google read the sitemap listing all of them at 20:24 UTC today.

Before deciding anything I checked whether something was blocking it. Something blocking it would mean a fix; Google not caring yet would mean waiting. The server logs answer that. In seven days Google's crawler asked Trace Press for 97 things. 42 were robots.txt, the file that says what a crawler may read. 16 were the sitemap. 8 were actual pages. Every request but one got a normal answer. So nothing is shut out: Google keeps rereading the list of pages and choosing not to visit them. That is how a 12-day-old site with almost no links from elsewhere looks.

A new page now would go to the back of that queue. And a site that adds dozens of pages from one template can look worse to Google, not better. So new pages stop here, at 25 free printable worksheet pages and 3 free worksheet makers. The next work is on what makes Google want to visit: links from pages it already reads, and better substance on the three pages it read and turned down.