Performance
Fast pages come from the architecture: stories are built once and served as files. Here is what our tests measure today — and what they do not.
Measured in our automated tests
| What | Result | Where |
|---|---|---|
| Database queries while serving published articles, sitemaps, feeds and redirects | 0 per request (the test fails otherwise) | Docker test environment, OpenResty edge, MariaDB server counters |
| Published page when the database, cache and application are stopped | Still served | Same environment (outage test) |
| Importing a WordPress export of 1000 posts and 333 images | about 150 s including image processing; a second run changes nothing | Developer workstation (Windows 11), single worker |
Measured on the pilot server (October 2026)
| What | Result | Where |
|---|---|---|
| Migrating a real WordPress news site: 2,000 posts, 1,695 images, 8,651 tags, with redirects | 27 minutes until every page was live (import 16 minutes, then publishing) | Shared Hetzner server (also running other services), three publishing workers |
| Rebuilding every page of that site after a settings change (about 10,100 publishing jobs) | 111 seconds; CDN cache cleared with 5 requests | Same server |
Engineering targets (not guarantees)
Article pages: about 35 KB of first-party JavaScript or less, theme CSS of about 45 KB or less, LCP ≤ 2.5 s, INP ≤ 200 ms and CLS ≤ 0.1 at the 75th percentile of real visitors.
Not measured yet: real-visitor (field) data, CDN cache behaviour and response times from different regions. We will publish these with the network, region and device profile when they exist. Lighthouse scores are not a product promise.