The discovery algorithm is the right shape: properties first, type inferred. It means the authoring side stays tiny — just markdown frontmatter — and the reading side gets all the structure it needs.