What these numbers mean
The scores are generated from the production build, not chosen by hand. The public number is the lowest result across the homepage and a long, code-heavy article at Lighthouse’s mobile profile. A score of 90 or better passes the quality gate.
Lighthouse checks
- Performance: 99/100. Loading, rendering, main-thread work, and Core Web Vitals lab estimates.
- Accessibility: 100/100. Automated semantic, contrast, naming, landmark, and keyboard-related checks.
- SEO: 100/100. Indexability, canonical metadata, mobile rendering, links, and crawlable page structure.
- Best practices: 100/100. Browser safety, modern APIs, console errors, and implementation hygiene.
Lighthouse is a lab test, so performance can vary slightly between machines. Accessibility automation also cannot prove that a site is fully accessible. Keyboard behavior, reduced motion, focus restoration, responsive reflow, and content clarity are reviewed separately.
Site-wide structure
The audit also makes 8 structural assertions across all 9 generated HTML pages and all 4 articles. It requires a language declaration, unique page title, description, canonical URL, main landmark, H1, JSON-LD, and a Markdown alternate on every article. All 8 checks currently pass.
Accessibility review
Beyond Lighthouse, the shared layout provides a skip link and visible focus treatment. The mobile menu closes with
Escape and restores focus. Technology tabs implement arrow, Home, and End navigation. Filters and theme controls
expose pressed state. Motion-heavy effects are disabled under prefers-reduced-motion, and layouts are
checked down to a 320-pixel viewport.
Search and sharing
Every page has canonical, Open Graph, Twitter, author, and JSON-LD metadata. Articles add publication, modification,
tag, author, and BlogPosting data. The site also publishes a sitemap, robots policy, validated RSS feed,
and a 1200×630 PNG social image.
LLM-readiness checks
Lighthouse 13 checks whether llms.txt can be fetched. The
turva-llms-txt-validator package adds formal structure and v2 discovery validation.
The 100/100 score combines those independent checks
with the site's endpoint and content assertions:
20 of
20 checks pass.
- A concise /llms.txt index is available.
- A complete /llms-full.txt corpus is available.
- Every published note has a source-derived Markdown mirror.
- Article HTML advertises its Markdown alternate.
- Pages advertise the site-wide LLM index.
- Markdown mirrors preserve headings, links, and fenced code.
- MDX-only imports and presentation wrappers are removed.
- Canonical URLs and metadata accompany machine-readable notes.
- Robots and sitemap endpoints are available to crawlers.
- Structured data is valid JSON and identifies articles and their author.
Reproduce it
Run npm run audit:site with Chrome or Chromium installed. The command builds the site, starts the
production preview, audits both representative routes, verifies the machine-readable endpoints, writes the
measured results to src/data/site-quality.json, and rebuilds the displayed report.
Do not run Lighthouse against npm run dev. Astro's development mode deliberately serves the Vite
client, developer toolbar, source modules, and unoptimized responses. They are useful while editing, but none are
shipped by the static production build and including them does not measure this site.
Known limits
The large article produces a correspondingly large DOM because syntax highlighting wraps individual tokens. This is an authoring tradeoff rather than an interaction bottleneck. The site also ships a small Alpine runtime for theme, menu, tabs, and note filtering. The shared Basecoat stylesheet is broader than this site strictly needs. With a measured 99 performance score, both are monitored rather than removed merely to game a benchmark.