James Ide

Expo’s Series B lets us build + More developer success for our world-class customers + More products to unlock that success + More capacity to meet the demand for our products and services We are also excited to continue collaborating with other great companies, and for the Expo SDK and services to work well with even more libraries, products, and integrations from others. We care and work to be a great partner to great partners. I’m proud of what we’ve built, from what started in a room w/@ccheever and is now a global team over 60 people strong. This is an energizing time to be at @Expo!

Of course an app about caffeine is the first app to use Exact in production

The way to think about Exact is the agent-native way to write high-performance native apps. @ccheever *doesn’t* describe Exact as cross-platform in his launch post. In the agentic era every team is empowered to build apps that reach their users everywhere they are. Your agent can write N versions of your app with Exact or anything else for N platforms. When N is a large number of target platforms - and platforms add new intrinsic complexity like iPhone Duo - it is taxing to multiply your token cost by N. Every company is using agents to build faster on their traditional codebases. They are working on faster horses and, in some cases, N horses. Exact is the car. It is agent-optimized from the Contract programming language, context-rich errors, testability, and more @ccheever shares in his post. You can manage N codebases with Exact - N cars, if you will. However, Exact aims to bring the token cost of building for N platforms down from ~Nx to ~1x, or at least a fraction of ~Nx. In practice a lot of declarative UI code can be shared in a semantically correct way when the host platforms themselves share similarities. And Exact is built to Use the Platform through and through, except when the platform’s implementations hurt performance too much, in which case Exact uses lower-level native APIs (example: text rendering). To set expectations, a single-platform Exact codebase does not magically work perfectly on all other platforms. Each requires some platform-specific input to get exactly what you want. However, this first Exact codebase will give your agents a competitive lead to reach a previously ambitious number of platforms. Finally, agent-optimized code generation with Exact slots into the rest of the Expo agentic infrastructure to help developers achieve autonomously improving apps. Expo + Exact = self-driving car for application development

1/ @ccheever deserves most of the credit for Exact. He has been working very hard on it and making Claude/Codex/Grok/Muse work very hard on it. 2/ To get started point your agent at github.com/expo/exact and start by building a few screens. Pro tip: tell your agent it can edit and compile Exact if it needs to. Agents make local patches maintainable, do what you need to stay unblocked. 3/ Tell your agent to share feedback, both what went smoothly and where there was friction or bugs, by enabling agent diaries or posting an issue on the GitHub repo. Thank you!

Contract is a new agent-first programming language for writing high-performance native UIs. It’s a key part of what makes Exact the agent-native way to write pre-agentic-era native frameworks like UIKit on iOS, AppKit on macOS, SwiftUI on Apple Watch, Android Views & Canvas on Android, the DOM on web, and so on. Contract is fast to write, fast to deploy, fast to run. The language is designed for agents to write. Deployment is extremely fast with hot reloading and over-the-air updates to get feedback in 10s to 100s of milliseconds. Contract bytecode runs on Rust and uses fine-grained updates to minimize how much of the UI is updated. As a human metric, an iOS app written with Exact often renders before its window fully expands to fill the screen when the user launches it by tapping its home screen icon, a metric I call “Rendered Before Zoomed”. Contract is an evolving language. This takes advantage of agents’ tremendous ability to “learn” new programming languages, and the fact that Contract is already fast for agents to write despite its not being in the model weights yet. Contract will become an even more token-efficient way to use native frameworks as models train on it.

Let Siri use your app with App Intents today. Livestream now.

iPhone Duo as a Service Have an iOS app but no Duo? Reply with a link to your app and I’ll film a video of it on a physical Duo. Steps so simple you don’t even need an agent. Run: npx appdrop <devicebuildofyourapp.ipa> Reply to this post with the AppDrop link

Concerning how often compute is a bottleneck even with the latest silicon

Recursive self-improvement is when a public company’s agent has a connector for brokerage accounts but it can trade only one stock and there’s only a Buy button @SpaceXSI has the chance to do the funniest thing

I ❤️ Des- ign

Software: “UIKit size classes will automatically adapt to Duo” Hardware:

IEEE 754: NaN != NaN iOS 27: English (US) != English (US)

Gonna build computer use for phones so my agent can type on this honking keyboard for me 🫱⌨️🫲

Two things YouTubers w/early Duo units missed IMO 1\ How fast developers will be ready 2\ The (un)importance of ProRes/ProRAW Different contexts but apps will support Duo 10x faster than they did Dark Mode. And normal people haven’t even heard of ProRes, let alone use it.

Going to Tokyo next month to try floopy