Between 2026-10-09 and 2026-10-19 we shipped 231 commits across five repos, touching 1089 files, adding 31779 lines and removing 4597. Commits landed on 11 days. Two repos carried almost all of it: skat-game with 100 commits and new-Orqestra, the repo for Orqestra, with 99. MythMarket-Quantum had 28, and FitGro and NeuraGrowthHTML had 2 each.
In the same stretch two finished books joined the catalogue, both study guides for Claude certification exams. One kids film slot was added to the weekly schedule. And one operation in our agent system cost more than any other over the last 10 days.
This post goes through those pieces and what they say about how NeuraGrowth decides what to build. It is a brand-pillar log, so the point is the decisions, not a changelog.
Two Claude study guides in six days
Product 219, the Claude Certified Developer Foundations (CCDV-F) Study Guide, shipped on 2026-10-09. Product 220, the Claude Certified Architect Professional (CCAR-P) Study Guide, shipped on 2026-10-14. Both are books, and both are on sale at kdp and kdp_print, which means an ebook and a print edition.
The professional guide has its own page in our catalogue: https://neuragrowth.co/books/ccar-p/. It is listed there as a 407-page study guide.
Both are guides, which is where most of what we make sits. Two Claude-related books in a week is consistent with that, not a change of direction.
What a Claude-related book costs to plan
The most expensive operation over the last 10 days was kdp.plan.currency. It cost 14.74 dollars out of 141.20 dollars across all operations, which is 10.4 percent. It made 14 calls, averaging 1.0529 dollars per call, and every call used claude-opus-5-5.
The facts give us the operation name and the numbers. They do not say what the operation did inside the planning step, and we are not going to guess at it here. What we can say is that the single costliest item in a 141.20 dollar window was a planning operation with a KDP label, not a render, not an image job, not a publish step.
We are logging the cost because it is a technical fact about running the studio, the same kind as the price on a product's shelf. Fourteen calls at about a dollar each is a number we can look at again next time.
A second Wednesday film for the kids line
On 13.10.2026 Robert asked for a second film in the middle of the week. The change shipped on 2026-10-14 in new-Orqestra: a beat entry named kids-publish-wednesday that sends a bajka at 07:00 UTC every Wednesday, next to the 14:00 short. On the board it is named "Środowy film Kids".
The entry is what turns the slot on. It touched 17 files, with 111 lines added and 63 removed. A no-deploy window of 06:50 to 07:20 UTC on Wednesday now exists in deploy.sh, the same way Sunday's film has one. The tests that pinned a week of six shorts and one film now read eight slots.
One detail from the commit is worth keeping. Beat runs in UTC, so on 25.10.2026, when the clocks change in Poland, the two film slots come an hour earlier in Polish time, at 15:00 and 08:00, not later. The comment above the publish entries says so. We would rather write that down in the file than rediscover it in October.
The review that found what the first change missed
A follow-up change shipped the same day, 2026-10-14, as the review of Task 10. It was small: 7 files, 53 lines added, 19 removed. It fixed several things the first change had left open.
The health list now says the two Wednesday slots hide each other's miss. The film's 07:00 run also hides a missed Wednesday short from the 32 hour tolerance. A night test now checks that the Wednesday film is made when its slot is the earliest gap. The calendar test checks every silent slot of a day, where before a dict by day checked only Wednesday's last row.
Two more catches are worth naming. A Wednesday window typed as Thursday's passed the window test, so that test now checks each window's weekday. And the beat chip read "niedz. 09:00; śr. 09:00", so the Wednesday entry now comes before Sunday's and the chip reads the week in order. The commit also says that nothing tested the first thing production will do, which is why the night test exists.
None of this is glamorous, but it is where a schedule change either holds or quietly breaks. The first commit turned the slot on. The second made sure a miss would be visible.
How these pieces fit the way we decide
A Wednesday film was a direct request, so it got built within a day. The Claude study guides sit on the guide side, which is the biggest part of the catalogue.
Some earlier entries in this log cover the machinery around the products, for example what our Etsy review step looks like inside an agent system: https://neuragrowth.co/blog/what-our-etsy-review-step-looks-like-inside-an-agent-system/. The crontab side of a second weekly film also has its own entry: https://neuragrowth.co/log/a-crontab-entry-for-a-second-weekly-film-and-the-utc-hour-it-hides/.
When we add a scheduled slot, the work is not finished when the entry turns it on. Write the test for the first thing production will do, check each window against its own weekday, and make sure a missed slot shows up somewhere you will look. And when one operation tops the cost list, as kdp.plan.currency did at 14.74 dollars of 141.20, write down its call count and per-call price while the numbers are fresh, so the next planning run for a Claude-related guide has something to be compared against.
Originally published at neuragrowth.co. NeuraGrowth is a one-person digital-products studio; this is the log of what its pipeline does and where it breaks.
Top comments (0)