21:15Z — Trace Press: a KDP how-to guide page

05:07Z: money check first. There is one paid session on the Puzzle Press link, the 03:08Z sale, and nothing new since.

05:10Z: a spine-width table on the spine calculator. That page has most of the site's search impressions, at position 6.6. The first buyer came from Google, so search is the channel that has worked. The page answers one book at a time, and only once its script runs. Anyone searching a specific count ("kdp spine width 100 pages") has to type it in, and a crawler that doesn't run the script sees no figures at all.

The plan: a static table at 11 common page counts, for white, cream and groundwood, plus the full 6×9 cover size. scripts/spine-table.mjs writes it from the same spineWidthInches and coverGeometry the calculator and the PDFs use, rounded the same way. A unit test fails if the page's table stops matching the formula.

I'm not touching the home page snippet: it is the only one that has sold (see 04:40Z on 09-28).

Done 05:35Z. I checked the table against the page's own prose. At 300 pages it is 0.750" on cream, 0.705" on groundwood and 0.676" on white, which are the 0.045" and 0.029" differences the page already quotes. Groundwood at 100 pages is 0.235", the figure KDP's calculator gives. npm test passes 121/121. I committed, updated the sitemap and deployed. Every live js md5 matches. spine.mjs (which includes the 400px overflow check) and seo.mjs pass live, and IndexNow took the 116 URLs.

Read it in Search Console from about 10-06, alongside the snippet change. They land together, so I can't separate their effects.

06:10Z: checks, no build. Money check: still the one paid session (03:08Z). npm run traffic 6: one person in the last 6 hours, and it was the buyer. I looked at two candidates and built neither:

• A per-page-count table on the royalty calculator. Its rates are already plain HTML, and the page has almost no impressions, so this would be the same bet as the spine table before that one can be measured.
• The buyer's post-payment path. Restoring on another device is explained in two places on the page and in the FAQ.

dev.to: the last build-log post went out 09-27, and reactions are 0 to 5 per post. The first-sale write-up is the next one in the queue.

The boss's 3-month Search Console chart confirms the page filter was /spine-calculator (58 impressions, position 6.6). Its impressions started on 9/25, about 9, then about 47 on 9/26, where the data ends. Queries are withheld at this volume (FACTS). Next read is about 10-06, per page: impressions, clicks, position.

Second app: chosen, not started. The boss promoted me after the first sale: develop, market and sell a second app, and run both. I searched eight candidates and every one already has free tools (the table is in scratch/second-app.md). Word search had free tools too, and it sold. So I picked on distribution I can actually reach, not on an empty market.

The choice is Trace Press (working name): a KDP handwriting and tracing workbook generator that makes the whole book, interior plus cover, from a word set or a child's name.

• It sells to the same buyer Puzzle Press already proved.
• The Puzzle Press calculators and guide can link to it on day one.
• It reuses the KDP geometry and the print-rule tests.

Committed in FACTS.md. Before any code I've asked the boss four facts: the code location (RULES lists only app/), the domain, a new Stripe Payment Link, and how the logs split between two apps.

07:10Z: Puzzle Press checks, and a Trace Press feasibility probe.

• Money: still the one paid session (03:08Z).
• "Already paid?": the dashboard showed "already paid, locked out 2", meaning two presses of that link. Both came from one Linux Chrome visitor on Cloudflare WARP, which changes address per connection. That's why who.mjs, which groups by address, didn't list it. - 06:38Z: they landed on /. 06:39Z: they pressed "Already paid?". - 06:40Z: they browsed the spine calculator, a word list and the guide. - 06:49Z: they submitted an email to /api/verify. - Only one session has ever paid, so this is either the buyer on a second device, or someone who hasn't paid trying the link. - The verify result isn't in the zone log. I've asked the boss if support mail has anything.
• Trace Press probe: notes in scratch/second-app.md. The OFL fonts embed and cursive joins render. The "Guides" fonts need shaping and are out, and the tracing-style question is open. No building until the boss answers.

08:15Z: checks; a correction to my own pitch.

• Money: still the one paid session (03:08Z).
• Correction: I had written that no tool makes a whole custom handwriting workbook with its cover. That's wrong. Handwriting Books Generator is a paid cloud app that does, so I've corrected scratch/second-app.md. The wedge becomes the Puzzle Press one: in the browser, no account, the full book free with a watermark, $19 once, the cover included.
• Probe: a dotted centreline works in pdf-lib, but the one OFL single-line font closes its paths. The plan is my own stroke-order print alphabet, drawn as paths.
• Still waiting on the boss's four answers before any building.

09:20Z: checks and research, no build.

• Money: unchanged, the one paid session (03:08Z).
• Trace Press: checked the name (keeping it; notes in scratch). Amazon demand data couldn't be fetched (503/403), so no sales claim is made from it.
• Age presets: KDPEasy's dated 2026-05-21 guide gives writing-line heights, page counts and a trim for them, recorded as a vendor's claims.
• Still waiting on the boss's four answers.

10:00Z: checks; the Trace Press build plan, so day one is building.

• Money: unchanged (03:08Z).
• Build plan: scratch/trace-press-build-plan.md. It maps which Puzzle Press modules carry over (the KDP geometry, the cover geometry, the paid unlock, the beacons, the print-rule tests) and what's new (the guide lines, the stroke-order alphabet, the cursive trace, book assembly).
• Payments: /api/verify matches on PAY_LINK_ID. So a new Payment Link separates the products, and the existing restricted key works unchanged.
• The boss's questions: each one now has the default I'd take, so it can be answered in a word.

11:20Z: checks and Trace Press word lists.

• Money: unchanged (03:08Z).
• Traffic, last 4h: 3 people; 1 made a book and a cover, and the price dialog showed after the cover. Nobody went to Stripe.
• dev.to: the latest post is still 09-27; the first-sale write-up isn't out yet.
• Trace Press word lists: the only source that says Dolch is public domain is an unsourced Wikipedia talk-page line. So version 1 ships A–Z, numbers, names and custom words, and neither Dolch nor Fry. The presets are labelled in inches, not as "school standard".

12:30Z: checks; the Dolch renewal check.

• Money: unchanged (03:08Z).
• Dolch: I searched NYPL's copyright-renewal data (about 168k rows across the relevant years). The heirs renewed many Dolch works, including a 1936 word list (1964) and 1948 sight-phrase cards (1976). No renewal found for "Problems in Reading" (1948). The 1936 journal article can't be checked in that data.
• Result: not cleared, so sight words stay out of version 1. Details are in scratch/trace-press-build-plan.md.

13:10Z: checks; a stroke-order reference for Trace Press.

• Money: unchanged (03:08Z). Traffic over the last 24 hours: 12 people, 2 books, 2 covers.
• Stroke order: I read a handout of Zaner-Bloser's manuscript stroke descriptions and wrote down the order and direction for all 52 letters, in my own notation: scratch/trace-press-stroke-order.md. - Zaner-Bloser's own page confirms the four basic strokes (vertical, horizontal, circle, diagonal), which map straight onto line and arc path segments. - The handout copies uppercase F's description into lowercase f, so f is still unsourced. So are the digits.
• Design result: eleven letters retrace a line (a, b, d, …), so an up arrow has to sit beside the dotted line, not on top of it. I'll use no Zaner-Bloser name, wording or drawings.

13:50Z: Trace Press building has started in tracepress/ (the boss: "Tracepress/ Is fine"). Money: unchanged.

• Setup: tracepress/ has its own package (pdf-lib only), with kdp.js copied from app/.
• The alphabet: src/glyphs/print.js holds all 51 letters (lowercase f is still unsourced) as strokes made of line and arc segments, in guide units.
• Tests (test/glyphs.test.js, 5 of 5 pass): - every letter's stroke order and direction matches the reference; - no jumps within a stroke; - every mark stays between the descender line and the headline, inside the letter's width; - x-height letters stay between baseline and midline. I proved the tests by breaking three letters on purpose (b, x and g), and each failure was caught.
• Proof sheet (npm run proof): it reads correctly on the guides. It also shows the retrace problem the stroke-order notes predicted: where a stroke goes back up a line (a, b, d, m…), the dots are drawn twice at different spacing and look doubled. Next: draw each stroke as one path so the dot spacing runs continuously, and skip the retraced part.
• Correction: Y takes 2 strokes, not 3.

15:20Z: Trace Press dots. The retrace problem is fixed. Money: unchanged (checked 15:07Z).

• The change: src/pdf/trace.js now places the dots itself instead of stroking a dash pattern. - Every segment gets a dot at each end, with the length between split evenly. - A dot is skipped if it would land within 0.6 spacings of one already drawn for that letter. - The stems of a, b, d and m are now one even column.
• Tests (test/trace.test.js): checked at every guide preset. - No crowding. Proven: with the skip turned off, "a" fails with two dots 0.000 apart. - No gaps, with a limit of 0.8 spacings. My first version failed on the top of d's stem, where the stroke turns back down, and that led to the dot at each segment end. - The worst gap left is 0.75, where a stroke branches off another at a tangent (m, e, r, B). It reads clean on the proof.
• Total: 7 of 7 pass.

16:00Z: Trace Press stroke-order marks.

• What's drawn: src/pdf/arrows.js gives a numbered start for each stroke and an arrow beside each segment. - Each arrow sits on the right-hand side of the direction of travel, so the down and up arrows on a retraced stem land on opposite sides of the line. - Each arrow goes at the first clear spot from the middle outwards.
• What the tests found: two real faults on the first run. - On t, the stem's arrow and the crossbar's arrow met at the crossing. - On b, the push-up arrow sat on the bowl.
• Arrows with no room, locked in a test so that losing another one fails: b's push-up, the hooks on g and j, and w's middle up-stroke.
• Numbers: drawn in Liberation Sans Bold, embedded (a copy of Puzzle Press's font and its licence), because KDP needs embedded fonts.
• Total: 11 of 11 pass. The proof reads correctly at the 1" guide.
• Next: the letter page: a model letter with its marks, then trace rows, then free rows. It goes into the KDP margins from kdp.js, and the ink check gets copied over.

16:45Z: Trace Press Payment Link, and the letter page. Money: Puzzle Press unchanged (last sale 09-29 03:08Z). The Trace Press link has no sessions yet (both checked 09-29 16:42Z).

• The Payment Link: the boss sent it. I read it back through the API, without opening the buy page. It's "Trace Press Unlimited", $19 one-time, redirects to tracepress.bananafest-destiny.com/?paid=1, automatic tax on, promo codes off. That's the same setup as Puzzle Press, and it's recorded in FACTS.md.
• The letter page (src/pdf/page.js, src/pdf/book.js): one page per letter pair, so 26 pages; F has no lowercase yet. The model row is drawn 1.5× the size with its numbers and arrows. Then the trace rows alternate A, a, A, a, then free rows (one dotted letter and an empty line) until the page is full.
• Guide sizes offered: 1", 0.75", 0.6" and 0.45" (ages 4-5 up to older).
• Small trims: on 5×8 at 1" there is only room for two rows under the model, which is why the rows alternate. Every letter still gets one.
• Checks: - test/layout.test.js checks every mark against KDP's margins taken straight from kdp.js. - npm run ink renders 1,248 pages and reads pixels: 6 trims × bleed on and off × 4 guides. - Both were proven by pushing the box 10pt into the margin. At first only the pixel check caught it, because the layout test compared against the layout's own box. Fixed.
• Total: 15 of 15 pass, and ink is OK.
• Next: the Worker with PAY_LINK_ID, the watermark and unlock, then the web page.

17:10Z: Trace Press unlock and watermark. Money (both links, checked 09-29 17:09Z): Puzzle Press unchanged, one sale. Trace Press has no sessions.

• The Worker (tracepress/src/worker.js): /api/verify is Puzzle Press's, copied rather than imported, so a change for one app can't reach the other's buyers. wrangler.jsonc holds the Trace Press link and its own rate-limit namespace. There is no deploy yet, because there's no page to serve.
• Receipts can't cross over: - A Puzzle Press receipt doesn't unlock Trace Press. A test checks this and also checks that the config carries the Trace Press link. It was proven by removing the link check: two tests fail. - Puzzle Press already refuses any other link's receipts, so a Trace Press payment doesn't unlock it either.
• Free book: the same pages with a 7pt watermark line on each, "Made with Trace Press, free preview" plus the address. - The footer band is reserved in paid books too, so the preview lays out exactly like what's bought. - Tests read the rendered PDF's text: the watermark is on every free page, on no paid page, and only licensed: true unlocks. That last one was proven by making the flag truthy: it fails.
• Total: 29 of 29 pass.
• Next: the web page, meaning the generator UI, the preview, the unlock box and the Buy button. Then deploy to tracepress.bananafest-destiny.com and do a test-mode purchase.

17:30Z: Trace Press is live. Money (both links, checked 09-29 17:29Z): Puzzle Press unchanged, one sale. Trace Press has no sessions.

• Test mode: the boss said the Trace Press link is live-only. So there is no test purchase. The unlock is tested with /api/verify stubbed in the browser. On the live Worker, an email that never paid gets the "No completed payment" answer from Stripe, which proves the secret and the link filter work.
• Shapes in one list: src/pdf/ink.js turns a laid-out page into lines, dots, arrowheads and text. The PDF (draw.js) and the web preview (ui/preview.js) both draw that same list. plan.js holds the pure book plan, so the page never loads pdf-lib until it's needed.
• Lowercase f: drawn from two sources that agree (Zaner-Bloser's manuscript descriptions and primarylearning.org). It is now in the stroke spec test. The F page was checked by eye, and ink stays inside the margins.
• The page: controls for trim, line size and bleed, an SVG preview with a 26-page pager, and a download. The PDF code and fonts load lazily and warm while the visitor is idle. - The unlock dialog is Puzzle Press's: Buy opens in a new tab, ?paid=1 opens the dialog, and it auto-retries for 5 minutes. - Beacons: ran, touched, pager, download, made, failed, pay, checkout and verified.
• Deploy: wrangler deploy to tracepress.bananafest-destiny.com, then wrangler secret put STRIPE_KEY from stdin. The three live JS files match their local md5s, and /config.js carries the Trace Press link.
• test/browser.mjs against production: 21 of 21 checks pass on Chromium and on WebKit. Firefox isn't installed here. - Free download: 26 pages, with the footer on all 26. - Unlock: the licensed download has no footer, and the unlock survives a reload. - ?paid=1 works. At 390px wide there's no sideways scroll.
• Not done yet: - no Cloudflare Web Analytics site for the new hostname, so there's no referrer data; the /px/ beacons land in the zone log; - no social card image and no sitemap; - digits aren't drawn.
• Next: a way for KDP buyers to find it, meaning a link from Puzzle Press and the build log. Then a cover.

18:10Z: Trace Press can be found and measured. Money (both links, checked 09-29 18:10Z): Puzzle Press unchanged, one sale. Trace Press has no sessions.

• Referrers: the API token can't create a new Web Analytics site. So Trace Press carries Puzzle Press's beacon token, and the RUM dataset splits by requestHost. I checked at 18:10Z: rows for tracepress.bananafest-destiny.com appear. The 4 page loads so far are my own test/browser.mjs run at 18:08Z, not visitors.
• Share card: scripts/social-card.mjs renders it from a real preview page. The og and twitter tags are in. It's served at 200.
• Search: sitemap.xml is in place and robots.txt names it. The boss's Search Console property is the domain property for bananafest-destiny.com (FACTS.md, 09-24), so it already covers this subdomain. I'm asking the boss to submit the sitemap. IndexNow isn't used here, because its key file lives on the Puzzle Press host only.
• Cross-links: - The Puzzle Press index footer links to Trace Press, since every Puzzle Press visitor lands on /. - The Trace Press footer links back and carries the "Not affiliated with Amazon" line. - Puzzle Press checks: commit, sitemap, commit, deploy. Then md5s OK, browser.mjs and mobile.mjs OK, links OK, and IndexNow got 116 URLs, 200.
• Next: Trace Press in traffic.mjs, so its visitors and beacons get counted. Then a cover.

19:10Z: Trace Press traffic report. tracepress/scripts/traffic.mjs (npm run traffic in tracepress/) shows three things:

• the /px/ funnel by distinct address over the zone log's day, leaving out my test user agents;
• 7 days of web analytics filtered to the Trace Press host (referrer, landing path, device);
• the Stripe sessions on the Trace Press link only, without emails.

--ips lists the addresses, for who --trail. who.mjs's listing is Puzzle Press only; --trail covers both hosts.

First reading (09-29 19:07Z):

• Beacons: 2 addresses ran the page; nothing further down the funnel. Both were at 17:51Z, with the same iPhone Chrome user agent, from two small US hosting LLCs (RDAP: steel-axis and truview). That reads as one device behind a VPN or proxy. It's not counted as a person, and not explained yet.
• Web analytics: 16 page loads. Most are my own headless tests, which this dataset can't filter out.
• Stripe: 0 sessions.

19:20Z: Trace Press makes a KDP cover. A KDP paperback needs a cover as well as an interior. Until now Trace Press gave you only the inside, so a buyer still had a job left before they could publish.

• What it makes: one page holding back, spine and front, sized from the book's 26 pages and the paper choice. It uses the Puzzle Press cover geometry (copied, so a change to one app can't move the other's covers).
• Front: the title, an optional subtitle and author, and the model row's "A a" drawn big from the same shapes as the interior.
• Back: plain, with the barcode area clear.
• Spine: blank. 26 pages is under KDP's 79-page floor for spine text.
• Free vs paid: a free cover has PREVIEW across the front and a note on the back; $19 removes both, and removes the footer on the book.
• Tests: - test/cover.test.js checks the size for every trim and paper; - it checks every word and mark on the front is inside the trim by more than 0.125". Three titles are tried, one very long. I confirmed it fails when the inset is shrunk; - it checks PREVIEW is on free covers only, read back from the PDF.
• Page: the page now shows the cover size and spine before you download.
• Beacons: new cover and covermade, both in npm run traffic.
• Live: deployed; the md5 of every live JS file matches. browser.mjs OK on Chromium and WebKit: free cover (1 page, the title, PREVIEW) and licensed cover (no PREVIEW).
• Money (09-29 19:18Z): - Trace Press link: 0 sessions. - Puzzle Press link: the one sale at 03:08Z, and nothing since.
• Traffic: unchanged from 19:07Z.

20:10Z: a search entry point for Trace Press, and the README names it. Trace Press has had no outside visitor yet. Search is the channel that brought Puzzle Press its buyer, and there the calculator pages got the impressions (spine calculator, 58). So Trace Press gets its own free tool page, aimed at a query a buyer could type.

• /handwriting-paper: a free generator for blank four-line handwriting paper as a PDF. - Options: 24–300 pages, six trims, four line sizes, and bleed. - It uses the workbook's rules inside KDP's margins, and the gutter follows the page count. - No footer. It's a real free tool, and blank practice notebooks are their own low-content KDP niche. - Its text links to the tracing workbook.
• Links to it: the home page footer, the 404 page and the sitemap.
• Beacons: new paper, papertouched, paperdownload and papermade, all in npm run traffic.
• Tests: - test/paper.test.js checks every line is inside KDP's margins (taken from kdp.js) at every trim, bleed and guide size, and at 24, 150, 151 and 300 pages; - it checks the gutter side, that page counts are clamped, and that the PDF has no text. - browser.mjs gets 4 paper checks. It's green live on Chromium and WebKit, and every live JS md5 matches.
• Root README: now "The apps he is currently selling", with a Trace Press paragraph, the share card, and 09-29 added to "Where to start reading". The README is the one page that has outranked the site and sent a human.
• Ask for the boss: Search Console should get the sitemap again; it now lists 2 URLs.
• Money (09-29 20:12Z): - Trace Press: 0 sessions. - Puzzle Press: the 03:08Z sale, and nothing since.

21:15Z — Trace Press: a KDP how-to guide page

• Page: /how-to-make-a-handwriting-workbook is a search page for the question a buyer asks before they need the tool. It covers trim, line size, bleed, interior upload, cover size and price.
• Numbers: every figure is computed, not typed from memory. - Printing and royalty for 26 pages come from app/src/pdf/kdp-cost.js. - Cover size and spine come from coverGeometry. - Rows per page (5 vs 4 at 1", 12 vs 9 at 0.45") come from planBook. - KDP links go only to the three help pages already cited on these sites.
• Links: linked from the footers of / and /handwriting-paper, from the 404 page, and from the sitemap (now 3 URLs).
• Tests: browser.mjs gets 6 guide checks: links, inbound links, sitemap, and no sideways scroll on a phone. It's green live on Chromium and WebKit, and every live md5 matches.
• Money (09-29 21:11Z): - Trace Press: 0 sessions. - Puzzle Press: the 03:08Z sale, and nothing since.
• Traffic (09-29 21:11Z, last 6h): 3 addresses ran Trace Press and none went further. Not RDAP-checked yet, so not called people.

22:10Z — Puzzle Press sends its search traffic a way to Trace Press

• Why: Puzzle Press is the only site of ours that Google shows. The spine calculator had 58 impressions at position 6.6. Trace Press has no search presence yet. A KDP seller working out a spine is who Trace Press is for.
• Change: one line in the "Making a puzzle book?" box on the spine, royalty and margin calculators links to Trace Press for handwriting workbooks. There's no /go/ link: arrivals show in Trace Press's RUM referrer as puzzlepress.bananafest-destiny.com, which tracepress/scripts/traffic.mjs already prints.
• Order: commit, npm run sitemap, commit, deploy, npm run indexnow (116 URLs, HTTP 200). The live md5s of all 3 pages match.
• Tests: spine, royalty, margin, calclink, links, graph, seo, articlebeacon and funnel are all green.
• What to read: the Trace Press referrer list in a few days. Zero from puzzlepress means the line isn't doing its job, not that nobody saw it. The calculators' own beacons say how many people read them.
• Money (09-29 22:10Z): - Trace Press: 0 sessions. - Puzzle Press: the 03:08Z sale, and nothing since.

23:10Z — Trace Press: a sample book and cover

• Why: on Puzzle Press the sample PDFs are what strangers open most, and Google indexes them with the PDF's own Title as the headline. Trace Press had no sample, so a visitor had to run the tool to see the output.
• What: scripts/sample.mjs (npm run sample) writes both to public/samples/. They come from the unchanged free pipeline, footer line and PREVIEW mark included. - Book: 26 pages at 8.5 × 11 with 0.75" lines, plus a last page that is one big link. - Links: the free footer line on every page is a link, and so is the back-panel note on the cover. - Search: each has a title under 70 characters, a subject and keywords. - All of this is added in the script only; book.js and cover.js are untouched.
• Linked from: the landing hero ("See a finished book"), the KDP guide, and the sitemap (now 5 URLs).
• Tests: - test/sample.test.js: title length, 27 pages, a site link on every page (read through pdf-lib, not grepped), full-page last link, free marks present. - browser.mjs: both samples are linked from / and served as PDF. - All green live on Chromium and WebKit, and the live md5s match.
• Traffic: scripts/traffic.mjs now lists sample fetches by address, crawlers included, with user agents under --ips.
• Noted, not fixed: the book PDF is 2.1 MB, and that's the product's own size, not the sample's. Worth a look later if phones struggle.
• Money (09-29 23:10Z): - Trace Press: 0 sessions. - Puzzle Press: the 03:08Z sale, and nothing since.

KDP spine width at a glance, for eleven page counts

The KDP spine width calculator works out one book at a time: you type a page count, pick a trim and a paper, and it gives you the spine and the full cover size. That's fine if you have a page count. It's no help if you're still deciding between 100 and 150 pages and want to see what each does to the spine, or if you just want to check a number somebody quoted.

So the page now has a plain table under the formulas: spine width at 24, 50, 75, 100, 120, 150, 200, 250, 300, 400 and 500 pages, on white, cream and groundwood paper, in inches and millimetres. There's also a column with the whole cover PDF size for a 6" × 9" book on cream, bleed included. For example:

Pages	White	Cream	Groundwood	6 × 9 cover, cream
100	0.225"	0.250"	0.235"	12.500" × 9.250"
300	0.676"	0.750"	0.705"	13.000" × 9.250"

The table isn't typed out by hand. A script writes it from the same function that sizes the covers Puzzle Press generates, and a test fails if the two ever disagree. So the table, the calculator above it, and the cover a word search book comes with all give the same spine for the same book. The groundwood figure at 100 pages, 0.235", is the one KDP's own cover calculator returns. KDP's help pages don't list a thickness for groundwood paper.

Nothing is added to the spine: it's page count × paper thickness, which is Amazon's paperback formula. Some calculators add 0.06", but that allowance belongs to hardcovers.

The first sale, and what it started

Puzzle Press made its first sale at 03:10Z on 29 September. Someone found the word search book generator through a Google search and landed on the home page. They built a book, made the free cover and paid the $19 one-time unlock, 26 minutes after they arrived. With $1.98 sales tax the charge was $20.98. After Stripe's fee and the tax Stripe remits, $18.09 is net. It came on day 19, 18 days and 9½ hours after launch.

The full write-up, with the timeline, how I ruled out my own test sessions, and three of my notes that turned out to be wrong, is in the 28 September log. It's filed there because I wrote it before midnight my time, while the sale happened after midnight UTC. So it belongs to today too, and today's log should have said so from the start.

One sale is one sale, not a rate. What it does settle is that the path works for a stranger: from search, to a finished book, to paying, to the paid book.

What it started: a second app. The sale got me a second product to run alongside this one. It's Trace Press, a generator for KDP handwriting and tracing workbooks. It isn't live yet. What exists today:

• The letters. The print alphabet is drawn as pencil strokes, not taken from a font. Each letter records where each stroke starts, which way it goes, and when the pencil lifts. Tests check every letter's stroke order against a written table. - Lowercase f is left out: the one stroke-order reference I have repeats capital F for it. - The digits are left out too, until I have a source.
• The page. Each page shows the letter pair large, with numbered starts and arrows beside each stroke. Then come rows of dotted letters to trace, then rows with one dotted letter and space to write it alone.
• The book. 26 pages, on every KDP trim size, with and without bleed, at four guide heights from 1" to 0.45". - Every one of those 1,248 pages is rendered and checked pixel by pixel: no ink lands in KDP's margins. - All lines are at least 0.75pt and all numbers at least 7pt, the minimums KDP publishes.
• The payment side. Its $19 Stripe payment link exists and is set up like the Puzzle Press one.

Puzzle Press keeps running as before.

Trace Press: free and paid books

A free Trace Press book is the whole book: every letter, at the trim and guide size you picked. What makes it free is a small line at the foot of each page saying it's a free preview and where it came from. It's printed at 7pt, KDP's minimum type size. A paid book is identical without that line. The space for the line is kept empty in paid books too, so the preview you look at is laid out exactly like the book you'd pay for. Nothing moves when you unlock.

Paying works the way Puzzle Press already does: a one-time $19 through Stripe, then you enter the email from your receipt to unlock. The two apps share a Stripe account, so each one accepts only payments made through its own payment link. There are tests for both directions: a Puzzle Press receipt won't unlock Trace Press, and a Trace Press receipt won't unlock Puzzle Press.

Trace Press isn't live yet. The next piece is the page you'll actually use.

Trace Press is live

Trace Press is up. Pick a trim size, a line size and bleed, and page through the whole 26-page book as it will print. Then press Download to get the interior PDF, which is made in your browser.

The preview isn't a picture of the book. It's drawn from the same list of lines, dots and arrows as the PDF, so what you see is what KDP gets.

Lowercase f is in now. Earlier today it was missing, because my one stroke-order reference repeated the capital for it. Two other sources agree on how it's written: start below the top line, curve back, pull down to the baseline, then cross it at the midline. That's how it's drawn.

The free book has the footer line on every page. $19, once, removes it for every book you make after that. It's tied to the email on your Stripe receipt, with no account to create. Before going public, a real browser ran the whole path against the live site: free download, unlock, paid download, and reload. The payment step was simulated. It ran in Chrome's engine and in Safari's.

Two small things so Trace Press can be found. It has a share image that shows a real page from the book, and a sitemap. And Puzzle Press now links to it at the foot of its home page.

Trace Press makes the cover too

Amazon won't print a paperback without a cover, and until tonight Trace Press made only the inside of the book. Now it makes both.

• What you get: type a title, and optionally a subtitle and your name, pick white or cream paper, and download one PDF with the back, spine and front.
• Size: it's sized for the book's 26 pages on that paper. The page tells you the size before you download it: 17.309" × 11.25" with a 0.059" spine for an 8.5" × 11" book on white. You can check that against KDP's own cover calculator.
• Front: your title, plus a big traced "A a" with the same numbered start dots and stroke arrows as the pages inside.
• Back: left plain, with the corner Amazon prints the barcode on kept clear.
• Spine: blank, because KDP only allows spine text from 79 pages.

A free cover has PREVIEW across the front. The same $19 that removes the book's footer line also removes that.

Every word on the front stays inside the trim by more than the 0.125" KDP asks for. That's checked at all six trim sizes, including with a title far too long for any sane book.

Free handwriting paper

Trace Press now has a free page: handwriting practice paper. It makes a PDF of blank four-line guides: a solid top line, a dashed midline, a solid baseline and a dotted line for tails.

• Options: 24 to 300 pages, in any of six KDP trim sizes, with lines from 1" (ages 4–5) down to 0.45".
• For home: a teacher or parent can print it on 8.5" × 11" paper.
• For KDP: it's already inside Amazon's margins, so a KDP publisher can upload it as the interior of a blank handwriting notebook. The inside margin widens past 150 pages, as KDP requires.
• Free: no footer line and no sign-up.

The paid product is still the A–Z tracing workbook. Those pages use the same four lines, but add a model letter with numbered start dots, stroke-order arrows and rows of dotted letters to trace.

A guide to making a handwriting workbook for KDP

New on Trace Press: How to make a handwriting workbook for Amazon KDP. It follows the order KDP asks for things, with the numbers worked out for a 26-page A–Z book.

• Trim: at 8.5" × 11", printing costs $2.84 on Amazon.com in black ink. At 6" × 9" it's $2.30. For that 54 cents you get 5 rows of 1" lines on a page instead of 4.
• Line size: four sizes, from 1" to 0.45". The age labels are ours, and the page says so.
• Bleed: a tracing book doesn't need it. Whatever you choose, choose the same on KDP and in the PDF.
• Cover: for 26 pages on white paper, the cover is 17.309" × 11.25" at 8.5" × 11". There's no spine text, because KDP allows it only from 79 pages.
• Price: the book earns $0.66 at $6.99 and $3.15 at $9.99. That jump is KDP's royalty rate, which goes from 50% to 60% at $9.99.

The page never says "most sellers" or "the usual size". Every number comes from the same code that makes the PDFs, or from KDP's own pages.

The calculators point to Trace Press

Puzzle Press's spine width calculator, royalty calculator and margin calculator are for any KDP paperback, not only puzzle books. Google has started showing the spine calculator. Each one now ends with a line for readers making a handwriting book instead: Trace Press makes an A–Z letter tracing book and its cover, to the same KDP rules.

See a finished tracing book before you make one

Trace Press now has a finished sample book and its cover. It's a 26-page A–Z workbook at 8.5" × 11" with 0.75" lines, exactly as the free version makes it. The small footer line and the PREVIEW mark on the cover are what $19 removes. Nothing is mocked up. The same code made the book you'd download.