Measured, not promised

performancephilosophy

Every performance claim on this site comes from a measurement we actually ran. That's the whole idea behind the phrase this post is named after: measured, not promised — we did the measuring, and you keep the ability to check.

Three habits make that real.

Every number carries its method. A figure on this site names what was measured, at what scale, on which engine — "Measured on Vue 3.5, 20,000 instances, one full read pass" — so you can repeat it, on your hardware, and see what you get. And when a sentence describes a behavior, that behavior was executed before the sentence was written: components mounted, watchers fired, heaps snapshotted. The method travels with the number, always.

The benchmarks run in your browser. The best machine to measure ivue on is yours — so the load-bearing numbers execute live, right here, on the same engine that ships to npm. Press the button and watch your own hardware produce them:

Create 100,000 instances, in your browserLive · runs the shipped engine

ivue · new Class()

·

reactive(new X())

·

composable factory

·

Same shape for all three: two state values and a derived area. Numbers depend on your machine. The ratio is the point.

The tools come with the claims. The full benchmark protocol is published. The import-cycle audit is seventy dependency-free lines you can point at your own codebase. The census scripts behind the discipline numbers are printed in the posts that cite them. Everything we measured, you can measure — including on your own project.

Numbers age; machines differ; engines improve. The method is the part that stays true — which is why every number here brings it along. Measured, not promised. That's all it means, and it's enough.