Ray Fernando describes extracting a specification from an existing app and using that document to guide a fresh implementation. The specification preserves intended features as a reusable recipe, allowing the implementation to change without abandoning the product decisions already made. The planned rewrite prepares the app to run AI models locally instead of relying on cloud processing.
Ray Fernando divides the rebuild into phases with explicit ways to verify completion. The intended workflow has a coding agent test each phase before advancing, making progress easier to assess than one large implementation request. The first phases have produced a working interface and demo, but microphone input still needs integration and local-model support remains a goal rather than a completed capability.
Watch on YouTube
