I’m working on an experimental media specification called Self Media Decoder (SMD).

The core idea is to move from a “codecs-in-OS” model to a “codecs-in-media” model: each media file can embed its own decoder logic as a WebAssembly module, alongside the actual media data. A player executes the decoder shipped with the content, rather than relying on pre-installed codecs. Some technical aspects that may be interesting to this community: • Decoder logic is versioned and can be hot-swapped mid-sequence (“highest version wins”), allowing upgrades without re-encoding existing content. • Decoder state is preserved across fragmented media and branching hierarchies (DAGs), enabling non-linear playback, multi-angle, or adaptive narratives without resetting context. • Targets modern Wasm features (SIMD, threads, memory64) for near-native performance.

This is early-stage and intended primarily as a technical exploration rather than a product pitch. I’d be very interested in feedback, prior art, or criticism, especially from people familiar with codecs, containers, or standards work (e.g. MPEG, WASI, streaming systems).