Cold start is not one number
A single TTFD hides process birth, class load, and first useful frame. Here is the three-slice chart we teach in Field Benchmark Mastery.
Read →Journal
Notes that started as arguments in the lab
These pieces are written for people who already open a profiler. They accompany the courses; they do not replace them.
When battery traces contradict frame time
A smooth scroll can still drain a cell. How we teach paired attribution without picking a favourite metric.
Read →
Modeling Thai carrier jitter in the lab
A fixture built from Bangkok commute captures, including the 5G-to-4G hop near the river.
Read →
Why lab scores drift after release
Build flavour, thermal history, and fixture version: the three quiet reasons a “green” gate fails in production.
Read →