Seven senior React Native engineers. One collective.
No bench, no juniors hidden behind a senior's name
Any one of us can carry your mission, and nothing ships without another senior's review
This is https://t.co/gWjdX48mws 👇
Seven senior React Native engineers. One collective.
No bench, no juniors hidden behind a senior's name
Any one of us can carry your mission, and nothing ships without another senior's review
This is https://t.co/gWjdX48mws 👇
Ça stream en ce moment même sur le Cross Platform Show
T'as encore le temps de nous rejoindre pour poser tes questions en live !
https://t.co/GbhN9Sbmu3
Meet Mistral Large 4, aka Le Chonk.
• 1T parameters, natively multimodal. 49B active.
It is the best open weights model from US or Europe on aggregated benchmarks.
• State-of-the-art on critical workloads, including cyber defense, manufacturing and finance and it surpasses closed frontier models on visual grounding.
• Forged in Europe end-to-end and is deployable from Europe via our own Mistral Cloud infrastructure.
• Available to all via API today. Working with cybersecurity partners privately.
Open weights release end of October.
Still young (0.17): in my runs the agent sometimes closes a bottom sheet without wrapping up the step
But mixing deterministic + agentic steps with a replay cache is a really smart idea !
Your e2e tests break every time you move a button
I tried e2e by @TesterArmy on my iOS app: describe a goal with agent.act(), keep locators where you need precision
Run 1: the agent figures it out. Run 2: replayed from cache, no AI
Renamed a tab: Maestro broke, e2e didn't
Open source (Apache-2.0): https://t.co/07VZi6IKQ8
Built by @okwasniewski, @SzymonRybczak and the rest of the TesterArmy team
On mobile it runs on agent-device by @callstackio
Congrats guys !
We are releasing React Native Skia version 3.0 which runs WebGPU by default. By moving away from the OpenGL-centric design, it allows for multithreading by default as well as deep integration with three.js, vision camera, and transformer.js
https://t.co/FUpMonlHd3
I made Intl.PluralRules, RelativeTimeFormat, ListFormat and Locale native on React Native's Hermes engine.
Same app, same tap, real time on a Pixel 8a: 89× faster than the FormatJS polyfills.
Up to 385× per call. 399 KB less JavaScript. One import.
hermes-intl, open source 👇
Same output on iOS and Android: one 2,656-case corpus, one hash on both. 838 of 842 test262 runs pass.
The trade-off: 1.0 to 1.7 MB of native code per ABI with all 514 locales, about 0.6 MB for a custom 5-locale build.
I made Intl.PluralRules, RelativeTimeFormat, ListFormat and Locale native on React Native's Hermes engine.
Same app, same tap, real time on a Pixel 8a: 89× faster than the FormatJS polyfills.
Up to 385× per call. 399 KB less JavaScript. One import.
hermes-intl, open source 👇
Per call on a Pixel 8a, vs FormatJS (median of 3 runs):
https://t.co/XmvFwmxHAG: 202×
PluralRules.selectRange: 385×
RelativeTimeFormat.format: 318×
first new Intl.Locale(): 126×
ListFormat.format: 8.5×, joining strings is already cheap in JS