The pipeline is deterministic: the same input produces an identical output, verified by content hashing rather than asserted. Every stage writes its input and output hashes into a provenance chain, and the chain is validated rather than merely recorded.
The immediate benefit is ordinary engineering. A bug you can reproduce is a bug you can fix, and a build you can reproduce is one you can ship from twice. Generators that quietly vary between runs make both of those harder than they need to be.
The second benefit shows up when an asset changes hands. A provenance chain answers what the asset was made from, which stages touched it, and whether anything was altered afterwards — which is the question a buyer, a licensor or a platform actually has, and one that a screenshot cannot answer.
The mesh mathematics is pure standard library. No numpy, no scipy, and the engine carries zero external runtime dependencies. That is unusual and it is deliberate: it is what makes the thing self-hostable and air-gappable without a dependency audit becoming a project of its own.
- Deterministic and hash-verified, not best-effort.
- SHA-256 provenance chain, validated at every stage.
- Zero external runtime dependencies in the engine.