AI-Written OOXML Viewer Renders Office Docs in Browser via Rust+WASM

A new open-source library called @silurus/ooxml renders DOCX, XLSX, and PPTX files directly in the browser using HTML Canvas, with parsers written in Rust and compiled to WebAssembly. The project’s headline claim is unusual: the author states that every line of code — Rust parsers, TypeScript renderers, tests, and tooling — was produced by Claude through iterative prompting, with no human-written application code in the repo.

The architecture splits work between a Web Worker that parses the OOXML archive via WASM and a main-thread renderer that draws to canvas. Rendering deliberately stays on the main thread so the canvas inherits the document’s FontFaceSet, avoiding the silent font fallbacks an OffscreenCanvas would introduce. Each format ships both a high-level viewer and a headless engine (DocxDocument, XlsxWorkbook, PptxPresentation) so callers can build custom UIs like thumbnail grids or master-detail panes. A separate ~3MB MathJax-based engine handles OMML equations as an opt-in entry point, designed to tree-shake away when unused.

The project is notable less for the viewer itself — browser-side Office rendering has existed for years — than as a substantial demonstration of AI-generated systems code spanning multiple languages, a WASM toolchain, and framework integrations for React, Vue, and Angular. It adds a concrete data point to the ongoing debate about what production-grade software written entirely by an LLM looks like in practice.