Snapp Technology
Hi folks,
We’ve got a very “system edges are getting interesting again” issue this week. App Intents are moving from cute demos into actual routes through Spotlight, Siri, and our app data. Swift Charts keeps proving it can do the weird product-polish stuff if we’re willing to understand the marks. And off to the side, Swift on Wasm, local AI engines, and SwiftUI in Final Cut Pro are all quietly saying the same thing: Swift is leaking into more places than our old mental model allowed.
Enjoy!
Articles
Swift
Swift for Wasm August 2026 Updates
Swift for WebAssembly had one of those months where the changelog looks small until you notice the shape of it: WasmKit running a WASI guest and GDB stub on an ESP32, BridgeJS getting cleaner module imports and generic imported functions, ElementaryUI tightening up its SwiftPM/Vite story. That’s not “hello world in the browser” territory anymore. It’s Swift being pushed into edge workers, embedded targets, and frontend experiments with fewer apologies. Still rough in places, sure, but if our idea of Swift stops at apps and server binaries, we’re missing where the interesting pressure is building.
AI/ML
Building AI features using Foundation Models. Multimodal input.
Foundation Models get much more real once the prompt can include an image, and the plate-scanner example makes the point without overcooking it. Define a Generable result type, pass instructions plus an Attachment into LanguageModelSession, and get structured output back that the UI can actually use. No giant cloud pipeline, no “AI strategy” meeting, just a feature-sized loop: image in, typed result out. That’s the kind of AI API surface I want in our apps — small enough to trust, useful enough to ship.
UI/UX
Teaching Siri to talk about your favorite walks in Walk Mate
A practical walk through making Siri understand app data, not just launch a shortcut and wave vaguely at the app. Walk Mate’s favorite routes become App Entities with readable properties, Spotlight indexing, Shortcuts search, and a Maps handoff, which is exactly the kind of plumbing our apps need if we want Apple Intelligence to feel useful instead of decorative. I like how grounded this is: one saved walk, one name Siri can say, one route that opens where the user expects. App Intents can get abstract fast; this keeps the work close to something a person would actually ask for.
Making app entities available in Spotlight
Apple’s guide lays out the official path for getting AppEntity types into Spotlight, with IndexedEntity doing a lot of the connective work. The big bit for our codebase is that Spotlight is no longer just a bag of searchable strings; it can carry structured app entities that Siri and other system experiences can reason about. That means the boring indexing layer suddenly matters a lot more. If our app has user-created objects, saved places, documents, workouts, recipes, whatever, this is the place to start making them visible to the system.
Building a sunburst diagram in Swift Charts
Swift Charts can build a proper sunburst diagram, but the fun is in the details: SectorMark rings, radius ratios, angular insets, grouped data layout, and accessibility labels that don’t get treated as an afterthought. The result is a multi-level donut chart where the outer rings still line up with their parent categories, which sounds simple until you try to fake it by hand. Who doesn’t like a chart that looks custom without becoming a pile of custom drawing code? This is a good reminder that Charts has more range than the usual bar-and-line dashboard screenshots suggest.
Area chart with a dimming layer up to the current point in time
A nice little Swift Charts recipe inspired by the iOS Weather UV chart: gradient area, matching line, current-time marker, and a dimming layer over the part of the day that has already passed. The technique is useful because it’s not just prettier data, it tells the user what’s behind them and what’s still ahead at a glance. Our apps often need exactly this kind of visual hint, the small layer that makes a chart feel alive instead of pasted in. Bonus points for keeping accessibility in the conversation when the area and line represent the same data.
093 - SwiftUI in FCP
SwiftUI views rendered as Final Cut Pro effects is the sort of sentence that makes you stop and check you read it right. Using SwiftUIFX, the experiment builds SwiftUI components and renders them frame by frame onto video footage, which is not the usual plugin path most of us have in our heads. It’s scrappy and exploratory in the best way. Even if our day job never touches Final Cut Pro, seeing SwiftUI escape the app window is a healthy reminder that our UI tools are more portable than we tend to assume.
Utils
NobodyWho
NobodyWho is pitching local AI inference across devices, with text, vision, speech-to-text, and text-to-speech models loaded from places like Hugging Face and called from Swift, React Native, Kotlin, Dart, Python, and more. The Swift sample is refreshingly direct: load a GGUF model, ask it something, await the completed response. For mobile teams, the appeal is obvious — private, offline-capable AI features without sending every user interaction across the network. Of course we still need to care about model size, latency, and battery, but this is the kind of tool that makes local-first AI feel less like a research project.