One product thesis, another demanding workload
Models, datasets, checkpoints, and scientific results have the same foundational need as media pipelines: exact source history and an honest record of derivation. The Dits AI pages explore that fit; they do not describe a separate available product.
The shared foundation
The local engine can store arbitrary exact bytes in chunked, content-addressed history. Future derivation records could link data, code, configuration, seeds, tools, and outputs. That shared model is more coherent than maintaining an AI-specific storage engine or a second protocol.
The hard boundary
Generic CDC often reuses little across diffuse tensor updates. Dits has no tensor-aware format, experiment tracker, model registry, orchestration layer, remote transfer, or validated AI benchmark corpus. Those are open research questions.