How this site is built

We say this site does what it tells you to do. That's worth nothing unless you can check it, so here's how it works and where the gaps are.

How it works

In: Markdown files with required fields at the top, plus five data files — the entities, the statistics ledger, the glossary, the platform matrix, and the site constants.

Out: static HTML, a sitemap, llms.txt, llms-full.txt, and a search index.

Generated rather than typed, so it can't go stale: the breadcrumbs, the "more in this section" blocks, the navigation, the glossary links on first mention of a term, the structured data, the checked date on every page, the sitemap with real modification dates, and the search index.

There's no RSS feed. A reference site isn't a blog and we took it out.

The twelve rules

Each one stops the build and writes nothing:

required fields present · check date not lapsed · whoever is named stands behind it and has a stated credential · a valid evidence grade with a written reason · every figure traceable to the ledger · complete ledger entries · internal links resolve · glossary terms exist · one top-level heading and no skipped levels · sections have pages under them and those link back up · AI-assisted content on money, health or legal subjects needs somebody other than the author to sign it off · anything called documented cites the documentation.

One page is blocked by the eleventh right now. It's rendered but not published — no sitemap, no search index, nothing links to it.

The markup

Every page emits one JSON-LD graph. The firm and the website are referenced by canonical identifiers shared across Maestro's other sites rather than redeclared here, and the articles, breadcrumbs and question blocks all point at those.

Snippet directives are set as permissively as they go, on purpose. Google documents that a page qualifies for AI Overviews and AI Mode if it's indexed and eligible to show with a snippet, so a tight max-snippet would cost us the AI eligibility along with the snippet.

Question headings are real answer blocks: a question, then a first sentence written to survive being quoted on its own. The site search ranks those blocks, which makes it a working test of whether we followed our own rule.

Speed

Static HTML. No JavaScript on content pages. Self-hosted woff2 fonts with the two critical faces preloaded. No third-party origins, no fonts from a CDN, no tag manager, and no cookie banner because there are no cookies.

What we haven't checked

A page claiming the site passes its own checks should be specific about which checks have actually run:

• Core Web Vitals are unmeasured. The setup makes good numbers likely. Likely isn't measured, and we tell you to conclude from real field data, which needs real visitors.
• The structured data has never been through Google's Rich Results Test. Our build validates the shape and that every identifier resolves inside the graph. It can't validate against Google.
• No ledger entry has been re-checked since we first retrieved it. Each one records the date we looked, and the next review re-checks them.
• One row of the platform matrix is blank and says so, rather than being filled in from memory.