BuildHive · case study
One file. Three formats. Fifty-one images.
A content pipeline that regenerates an entire product line in fifteen seconds — built for a business of one.
- 12.5s51 branded images, rendered from scratch
- 1.5s32-page product — Markdown, PDF and cover
- 1source file behind all of it
- 25tests that stop bad output shipping
The problem
A solo operator was launching a digital product with a Pinterest channel behind it. The channel needs about fifty images, each carrying a different headline and pointing at a different page. Every one has to be on brand, correctly sized, and linked with tracking that survives.
Fifty images at four minutes each in a design tool is over three hours of mechanical work — repeated in full every time a headline changes, a price moves, or a keyword turns out to be wrong. Which it did, three times.
The same words also had to appear in a 32-page paid document, on twelve web pages, and in the site's own hero. Four places, four chances to drift apart, and no way to tell when they had.
What I built
A single source file — the twelve emails the product is made of — and a pipeline that renders everything else from it.
One file, content/emails.yml, generates four things: the 32-page product as Markdown and PDF, twelve web pages with their Open Graph tags, 51 pin images at 1000 by 1500 pixels, and the compose window on the site's homepage.
Fix a typo once and it is fixed in the product a buyer downloads, on every web page, and in every image — in about fifteen seconds. There is no second copy of anything to forget.
The images are HTML
Rather than drawing them in code, the pin templates are HTML that imports the website's own stylesheet, screenshotted by a headless browser. The consequence matters more than the technique: the brand colours are literally the same file the site uses. Change one and the site, the pages and all fifty-one images move together. They cannot drift, because there is nothing to drift from.
It refuses to ship broken work
The build fails rather than producing something subtly wrong:
- Any image not exactly 1000×1500 — asserted on every file.
- Any headline still clipped at its minimum font size, rather than quietly cropping it.
- Anything rendered before the webfonts finished loading, which is how fifty images end up in fallback Times and nobody notices.
- A template variable used but never declared — the check that stops a customer receiving an email opening “Hi [NAME],”.
- A required legal disclosure missing from the generated document.
Nothing publishes without a human
Publishing reads a spreadsheet where a person changes draft to approved. Dry run is the default and --live is required every time. The caps — three posts per run, twenty minutes apart — are constants in the code that no config file can raise, because an automation that fires fifty posts in an hour is indistinguishable from spam and the account is the whole channel.
What it produced
| Output | Time |
|---|---|
| 51 pin images — five designs across twelve pages | 12.5s |
| 32-page product — Markdown, print PDF, cover image | 1.5s |
| 21-page website, 100/100 on accessibility | 1.0s |
| Posting queue with a tracked link for every image | instant |
The site scores 99–100 on performance and 100 on accessibility on mobile, serves in 0.15 seconds, and works entirely with JavaScript switched off. It costs nothing per month to run.
When keyword research showed three of the product's target phrases were contaminated — ranking alongside fraud scripts rather than business advice — rewriting the affected headlines, web pages and images took one edit and fifteen seconds. Before the pipeline, that finding would have cost most of a day and probably been ignored.
What this is worth
The three hours saved per revision is the obvious part. Revise weekly and that is roughly four working weeks a year you stop spending on mechanical reproduction — the build pays for itself somewhere in the first quarter and keeps paying after.
The larger return is that revising stops being expensive, so you actually do it. You act on what the data says instead of living with the version you already paid to make.
Content Pipeline Build
$6,000fixed price, two weeks
One source file to every asset you publish, on brand, regenerated in seconds. Built on your own brand and your own formats. You own the code and it runs on your machine — no subscription, no platform, nothing to be locked into.
- A working pipeline for your content, not a template.
- Validation that fails loudly rather than shipping subtly wrong work.
- A README written for whoever maintains it in six months.
- Two weeks of support after handover.
Founding rate — $4,500 The first two builds are $4,500, in exchange for a named testimonial and permission to publish the work as a case study like this one.