Law is an incrementally patched, distributed-authored, path-addressed corpus – its structure is convergent evolution with software, and computing it correctly requires a compiler (LawVM) before any semantic analysis.
Key Takeaways
Amendments are typed operations – replace, repeal, insert, renumber, text-replace – with a target address, action, payload, and authority provenance; this vocabulary is consistent across Finnish, UK, Estonian, and EU systems.
Law’s tree structure (parts, chapters, sections) is a paper serialization format; it actually operates as a graph via cross-references, conditional overrides, and cross-jurisdiction dependencies that ignore hierarchy boundaries.
Law has at least two independent time axes – enactment time and legal effect time – plus retroactive amendments that rewrite historical state and ultra-active provisions that survive repeal for past events.
Text-level amendments target substrings within leaf nodes, below tree resolution; replacing “Secretary of State” with “Minister” is a string operation that simultaneously reallocates statutory authority.
LawVM scopes explicitly to the text layer (what provisions literally say at a point in time) and defers normative state (meaning, obligations, applicability conditions) to downstream tools – the argument is that semantic models built on unverified text produced decades of broken legal ontology work.
