Version Control for Law

In brief

Software engineers solved a problem that law has lived with for centuries: how to let many hands change a shared body of text without losing track of what it meant, who changed it, and why. This essay borrows their answer. It reads legal interpretation through the mental model of Git, and it builds that reading on the interpretive canons of Planet Gibsonia (using the numbering of the Medium edition), set out in a companion piece published the same day, Canons of Constitutional Interpretation of Planet Gibsonia.

The mapping

• Enacted law is the main branch. Amendments are merges with version tags. Under the rule the canons call Versio Enactionis, every law is pinned to the version of the interpretive canons in force at the moment of its enactment, so its meaning cannot drift silently as interpretive fashions change.
• Judicial opinions are signed commits. Under the good-faith canon (Bona Fide Interpretatio), a reasoned opinion states its textual basis, addresses counter-readings, discloses ambiguities and cites its sources. An opinion that does none of this is, in the essay's words, "pretextual and not entitled to the weight of precedent."
• Precedent is commit history: merged commits with their reasoning attached. An opinion with no reasoning trail is "a commit with an empty message."
• Overruling is a force push, and it is gated. The overruling court must name the canon that was violated, supply the canonical re-reading, address reliance interests and reason textually. An overruling that skips these conditions is "itself void as precedent."
• Dissents are preserved branches (Dissensus Servandus). Minority readings stay in the permanent record and can be revived under stated conditions.
• Incorporation by reference is the submodule problem (Relatio Certa). A reference is frozen at enactment unless it is expressly dynamic, so an upstream change cannot quietly rewrite downstream law.
• Foreign law and moral philosophy are disallowed dependencies (Fontes Limitati). They work like an allowlist: outside sources may be consulted for illumination but are not binding imports.
• Typographical corrections are amended commits (Scriptoris Error), allowed only when the defect is shown from contemporaneous records, is clerical rather than substantive, and admits a single corrected reading.
• Conditional rules are feature flags (Manifestatio Ambiguitatis). The threshold conditions for a conditional canon must be declared publicly before it applies, like an audit-logged flag.
• Self-application is self-hosting. The canons govern their own interpretation.

Style guide or governance?

The essay's sharpest claim is a distinction. Scalia and Garner's Reading Law (2012), it argues, functions as a style guide: advice on how to read well. The Gibsonian canons are governance: a branching model, a commit policy, a signed-commit requirement, dependency pinning and protected-branch settings. A style guide tells a careful reader what to do. Governance tells every contributor what the repository will and will not accept.

The essay also reports a test. Run against landmark decisions, among them Dobbs, Heller, Bostock, Shelby County and Brown, the framework comes out party-neutral: it "does not track left or right. It tracks whether the opinion did its textual work."

The model is stated on the Institute site as the Version Control for Law framework, alongside the Gibsonian Canons, and several of its metaphors (the signed commit, the force-push overruling, the preserved branch) now serve as terms in the Lexicon.

This page is a summary, not the essay itself. Read the full Version Control for Law on Medium, where Ross Brodskiy first published it on 17 April 2026.