100 / 100 on Lighthouse.
Against this repo's own build, not a marketing deck.
Run today against bun run build && wrangler dev on this starter — the same worker runtime and the same brotli encoding production serves, on localhost so the number is the site and not your connection. Lighthouse 13.x, headless Chrome. Re-run it in the next five minutes.
Speed
Speed is a measurement. Not a claim.
This page times its own load in your browser while you read it. Transfer weight and the Lighthouse rings are not live — they are the last recorded audit, dated, measured on the build actually shipping. All of it is reproducible: re-run the command below against your own clone.
Lighthouse's harshest default: throttled mobile CPU + Slow 4G.
Same build, same page, two throttling presets — the gap is the network, not the code.
Live — your browser measured this
Lighthouse measured
Same page, unthrottled — what a broadband visitor actually sees.
How to re-run this yourself
Every number on this page comes from a Lighthouse CLI run against this repo's own production build — clone it and check.
What I'll be honest about
- ·These are lab scores, not field data. The figures above are medians: 8 desktop runs and 5 Slow 4G runs on one machine. Headless Chrome on a Slow 4G throttle profile is a consistent benchmark, not a guarantee of what real users on real networks will see. The spread is real — the first request to a cold preview server scored 83 on desktop where every warm run scored 99, so run it more than once.
- ·Use `wrangler dev`, not `astro preview`. The preview server sends the document UNCOMPRESSED — 421 KB where production sends 69.7 KB brotli. On a simulated Slow 4G link that 6x penalty costs about ten points, and it measures the preview server rather than the site. `wrangler dev` runs the same worker with the same encoding.
- ·These numbers are for the homepage only — pages with heavier islands (payments, wallet) will score differently. Run the command above against any route you care about.