Engineering · Benchmarks
Numbers we measured.
Benchmark pages are usually marketing wearing a lab coat. Ours has one rule: every number below was measured by us, on artifacts you can inspect - starting with this website. What we cannot publish honestly (client system internals), we do not publish at all.
The scroll film’s asset pipeline
- 2.1 MBmaster frame (4K PNG)
- ~48 KBstreamed frame (1920px WebP)
- 97.7%payload reduction
- 55 MBentire 914-frame film
The full story is in journal entry 001. The conversion pipeline is a ~60-line script (sharp, quality 70, 1920px) - the highest-leverage performance work on the whole site.
Site weight (measured on disk, pre-compression)
- Film engine - cinema.js 124 KB (GSAP bundled) + main.js 40 KB. Everything the cinematic homepage executes.
- Film styles - styles.css 84 KB, uncompiled, no framework.
- Editorial pages - pages.css 12 KB + 5-17 KB HTML each; the JS on a typical page is 4 KB (analytics prep) - interactive tools add 4 KB more. No framework, no hydration.
- Social card - og.jpg 68 KB; icons ≤36 KB.
- Caching - frames, JS and CSS ship with immutable, year-long cache headers; a returning visitor re-downloads none of it.
Operations
- 99.98% uptime across systems we operate - the figure our monitoring reports, published because we measure it.
- < 24 hours - reply time to every enquiry; a promise we publish and keep.
What we do not publish
Client API latencies, cold-start profiles and cloud bills are their confidential operational data - publishing them anonymised-but-vague would be exactly the decorative benchmarking this page exists to reject. What we can do: measure your system in a strategy call and show you the method.
Related: Engineering journal · Caching strategy · DevOps practices