Introducing https://t.co/ycxJEf1z7w!
A new space where I explore how the best apps in the world are built.
First piece:
How's Linear is so fast? a technical breakdown.
https://t.co/9Vu1syrn1i
Introducing https://t.co/ycxJEf1z7w!
A new space where I explore how the best apps in the world are built.
First piece:
How's Linear is so fast? a technical breakdown.
https://t.co/9Vu1syrn1i
Shipped a last-minute Expo SDK 56 feature: up to 2.8x faster codegen compilation on Android.
Enable it with `android.usePrecompiledHeaders: true` in expo-build-properties.
Next, I'll try upstreaming this to React Native so everyone benefits from faster builds.
Iโve been focused on this exact problem for a year. The issue as you state is o11y, upfront testing only gets you so far. Tracking ram, cpu, frame usage is a lagging indicator because release cycles are long. If you sample this from your user base in production itโs often too late.
Running e2e tests with perfetto measurements is one way to catch regressions, operationalizing them is expensive.
Iโve tried compiling Hermesโ from source to run targeted HBC performance tests and it proved to be useful. At the end of the day the major foot guns are: not memoizing Intl formatters, poor animation code, third party libraries cause a lot of toil. Happy to go into more detail if you want to reach out
@peterpme Took me a month of intense focus to grind through and at the end of it, it exposed every flaw the bridge was hiding and our over reliance on a single thread to do everything
@peterpme Using pnpm to hoist the modules saves space but it breaks building from source. Iโve been thinking about using sparse checkout but then youโre limited in testing end to end
Need to show LLM output in your RN app? ๐ค
Here's a fully native component from ๐๐๐๐๐-๐๐๐๐๐๐-๐๐๐๐๐๐๐๐-๐๐๐๐๐๐๐ ๐ with support for streaming Markdown.
Let us know if you find this useful for your project.