Stewart Code

Method · Under the hood

How this website is built

The short answer

How the pages and prices work

Every page on this site is plain, pre-built HTML, so it loads fast, while prices are read from a database on every visit and written into the page and its search markup at the same moment. That means a price change shows on the next visit with no rebuild, and the price Google reads always matches the price you see.

This page is for the curious owner, and for anyone wondering how a founder-led studio runs a site like this. Everything here can be checked: press "X-ray this page" in the header while you read.

The parts

Four layers, each kept simple

01

Static pages

Each page is written as one file and turned into finished HTML by a small build script, which also adds the shared header and footer, the structured data, the sitemaps and the AI summary files.

  • No page is assembled in your browser
  • No page builder or plug-ins
  • Every change is reviewed and versioned

02

Live prices

The catalog lives in a small database. A Cloudflare Worker reads it on each request and writes the current prices into the page and its Offer markup in one pass.

  • One database row feeds the page and the markup
  • Pages are marked so browsers re-check them, so nobody sees an old price
  • Hidden products disappear from public pages

03

Payments

Checkout runs on Stripe. The amount always comes from the database, never from the page, so it cannot be changed in the browser.

  • We never see or store card numbers
  • Clients manage billing through Stripe's own portal
  • Payment events update the client record automatically

04

The price editor

A private admin page where the owner adds products and changes prices.

  • Behind Cloudflare Access, open only to the owner's email
  • The server refuses any admin request that lacks Access's signed pass
  • Every change is logged: who, when, old value, new value

Tested, not hoped

How we know it works

The catalog, admin and checkout code ships with an automated test suite. Each test was first shown to fail against a deliberately planted bug, then to pass on the real code, so a test that cannot fail is never trusted. Examples of what the tests catch:

  • A hidden product leaking onto the public page, or a missing price showing as $0 instead of "Request a quote".
  • An admin change accepted without the sign-in pass.
  • A checkout created for a product that should only be quoted.
  • A payment event with a bad signature, or the same event applied twice.
  • The price on the page and the price in the search markup disagreeing.

A browser test then changes a price on the admin page and confirms the public page and its markup follow.

For search and AI

What machines read

PartWhat it does
Structured data (JSON-LD)Describes the business, the founder, each service, its prices and each FAQ in the vocabulary Google and AI assistants use. The FAQ markup is generated from the visible questions, so the two always match.
Sitemap indexOne sitemap per section, listing only the pages meant to be found, named in robots.txt.
/llms.txtA short, plain summary of the site written for AI assistants.
AI facts pageThe facts we want assistants to quote, each with the date it was checked.
/version.txtThe code version deployed right now. After each deploy we check the live site until this matches, then re-check the markup and sitemaps.

The X-ray

How "X-ray this page" works

The toggle in the header costs nothing until you press it: no extra code loads before then. When you do, it shows the page's heading outline, the structured data behind each section (including the price data behind a price card), the page's real load timings measured in your own browser, and the text an AI assistant would read. It is keyboard-accessible and respects your reduced-motion setting.

The full Site X-ray tool runs Google's PageSpeed Insights on any site you enter, side by side with this one: the four Lighthouse scores and real-user speed data when Google has it, plus a one-click link to Google's Rich Results Test for the markup.

Kept apart

Separate by design

This site has its own database, its own storage and its own Stripe products, tagged so its code only ever touches its own records. A test fails the build if the site's configuration is ever connected to anything outside its own resources. The same separation goes into client builds: your site, your data, your accounts.

FAQ

Questions about how this site is built

Why not use WordPress or a page builder?

For a site this size, pre-built pages are faster, have fewer moving parts to update and fewer weak spots. We do build on other platforms when a client needs them.

If prices come from a database, why are the pages still fast?

The page itself is finished HTML. The only live step is one small database read to fill in prices, done on the server before the page reaches you.

Can my site be built the same way?

Yes. The live-price catalog and admin editor are part of our web app work, and the speed and markup rules apply to every site we build.

Do you use AI to build this site?

Yes, as a tool. One senior operator plus a disciplined AI build system, checked by a separate AI review pass that hasn't seen the build, then by Shaun page by page. Every page still has to pass the same checks.

Next step

See where your site stands, then talk to the person who builds it.

Run the free X-ray on your own site, or book a free scoping call. Either way you see real numbers before you spend anything.