Vertical Slice: Speech-to-Markdown Editor, LLVM UI Codegen, and Landscape Split - in Our Own Language

We're building SMS - a statically-typed language that compiles to native ARM via LLVM. The runtime is ForgeRunner (C++ / Godot). The editor is ForgeStudio, which is itself written in SMS. Today we shipped a vertical slice that touches every layer of the stack simultaneously. That's the point of a vertical slice: prove that the whole column works, not just one tier. Here's what we built. 🎙️ Continuous Dictation in SMS The new SpeechRecognizer component looks like this in SML (our declarative UI language): SpeechRecognizer { id: mic language: "de-DE" mode: clean filters: "zdf,wdr,applaus,musik" } mode: clean activates post-processing: filler words, broadcast noise artifacts, and applause markers are stripped before the text reaches your app logic. The dictation flow works like this: mic.listen() starts recognition mic.result fires with a partial result SMS accumulates it into pendingText and immediately calls mic.listen() again - continuous dictation A confirmRow appears the moment dict