I wanted to try Exact from @ccheever@expo last night. I asked my agent to port a few screens from my Moovii app, and I’m really happy with the results already.
2 prompts later: iPhone, iPhone Duo, iPad, Mac, and web. All from one codebase 🤯
Our partner @thegameband's new game @dartwords is officially out on the App Store!
The game is full of thoughtful details. From the visuals and motion to sound design and voice acting, everything we built was intentional and done with care.
Play it now:
Maximize your chances of being featured by Apple by following @expo's guide to making your app optimized for Duo.
It mentions reserved-regions, which we've created and open-sourced while preparing our partners' apps. Simply follow the guide!
A line of people waiting to enter because the Wi-Fi drops isn't acceptable when Fortune 500 companies organize events.
Today we share how we built an offline-first system that has handled 3 million+ check-ins since launch for our partners @swoogo .
Link in reply.
Congratulations to our partners @fin for hitting such an important milestone! We helped build one of the most polished fin tech products on the market.
We’ve passed $1 billion in annualized payments.
We’re one of the few crypto companies that can say 100% of the businesses on Fin are non-crypto companies.
Safe area insets aren’t always enough on iPhone Duo.
Meet react-native-reserved-regions, a sister library to react-native-safe-area-context.
Apple’s new reserved regions API, in React Native. Let your UI make room for folds, camera cutouts, and system controls.
Android too.
Keep a toolbar clear of a camera cutout. Split content around a fold.
The API gives you region bounds relative to your provider view, so you can decide how your layout adapts.
While working on Stim, we had to make sure it actually brings value. To do this we built an evaluation suite and viewer. This allowed us to see exactly what an agent with Stim does vs one without.
You can play with it on our website: https://t.co/m73Khz3OtB
Some interesting findings:
- The different models perform similarly, so cheaper models are best.
- There is more variance for the agent not using Stim, as it tends to try different things.
- The token usage for the agent using Stim is a lot smaller since the output is optimized.