Raph Levien argues that Rust can support much faster interfaces, but a convincing demo leaves many hard problems unfinished. AI-assisted rewrites make shared foundations and clear specifications more valuable as new toolkits multiply.
- Reuse the difficult pieces. He points to Parley for text layout and AccessKit for accessibility, while warning that a superficial integration leaves important work undone.
- Give video its own layer. Letting the operating system combine video with interface controls can avoid expensive redraws. A widely adopted cross-platform solution remains missing.
- Separate design from execution. Define the interface between what an app should display and the engine that updates it. That gives both humans and coding agents smaller, testable modules.
For mobile, he considers platform-native interfaces with shared Rust application logic a viable approach. His performance argument is an architectural proposal, supported by examples rather than a controlled comparison of toolkits.
The discussion on Hacker News adds implementation experience and disagreement.
What the thread adds
- weinzierl — reports hitting the video-layer problem in Bevy: “ended up with frames copied into a texture and composited by the engine, because the app owns just one swapchain and the video can’t get its own layer.”
- skhameneh — describes the gap between smooth demos and a complete interface: “I have some nice markdown rendering and phenomenally fast smooth scrolling” but “I don’t even have text selection implemented yet”.
- raphlinus (Levien) — clarifies his architectural objection: “I do think you can go very far with an immediate mode GUI paradigm”. In that approach, the application describes its interface each frame. He argues that text caching and accessibility trees turn it into a hybrid.
__benny__counters that adding those features selectively can keep application state simpler. Their disagreement concerns where complexity belongs, beyond raw drawing speed. - nicoburns — answers an embedded-3D complaint: “My framework Dioxus Native allows this.” They link a working-example source tree. swiftcoder clarifies that they want a 3D view inside platform-native controls, rather than controls drawn inside a shared graphics surface. the__alchemist acknowledges: “That was indeed the part I’d missed.” That exchange separates 3D rendering support from the choice of surrounding controls.
- raincole and odo1242 — disagree about ArtCraft’s responsiveness. The former calls its apps “incredibly snappy”; the latter reports “lag spikes whenever e.g. moving clips on a timeline”. These are conflicting user reports, not comparable measurements.
HN handles are pseudonymous and have no public per-comment scores; ordering is HN’s ranking, so this is a slice of the thread, not a consensus.