How do we represent law so that it can be versioned, queried and audited?

Law already behaves like a versioned codebase. Enacted text is the main branch. Amendments are merges. Reasoned opinions are commits that carry their reasoning as a message. Dissents are branches the record keeps. Overruling rewrites history. What the law lacks is the governance a software team would demand of any codebase it depends on: signed commits, protected branches, pinned dependencies and a rule for when a force push is allowed. This program asks how legal knowledge must be represented so that it can be versioned, queried and audited, and it treats representation as the place where reliability is won or lost.

Version Control for Law supplies the model, and the Gibsonian Canons supply its governance: a closed, priority-ordered set of interpretive canons, written for a fictional jurisdiction so the design can be examined before any substantive politics begins, under which every law is pinned to the version of the canons in force when it was enacted. The Twelve Bridges address a different layer: the recognized doctrinal moves (direct holding application, analogical extension, a fortiori reasoning and the rest) by which a lawyer may lawfully travel from precedent to a conclusion, together with the fault patterns that mark an inference no court would accept. That matters because the most dangerous errors are no longer fabricated cases. They are inferential hallucinations: real authorities joined by invalid reasoning, which no citation database can catch.

Why this matters: retrieval over documents cannot see what a document-level index never stored. The Promise Fulfilled names five structural failures of that architecture (authority-health blindness, jurisdictional conflation, doctrinal boundary collapse, directional opacity and negative-space ignorance) and argues that three of them survive any amount of agentic searching, because they are failures of representation rather than of coverage. Propositions, treatment and inference have to exist as first-class objects before anyone can audit them. The computational jurisprudence tradition specified long ago what that representation requires. The work now is to build it at corpus scale and test it in the open.