Describe Glade integration and canonical model links

Clarify three-run Watch strategy measurements

Use qualified default specialization for Watch Core ML

Add experimental watchOS Core ML bundle and measured loading profile

Identify the Core AI conversion as a quantized derivative

Use Hub file checksums instead of redundant publisher manifests

Qualify iPhone quality shapes on ANE and publish clean measurements

Document measured iPhone performance, loading and mixed quality placement

Publish qualified 45-second and four-minute quality bundle

Document observed specialization and cached preparation

Publish independently measured fast and five-minute quality Core AI bundles

initial commit