Version Control for Law

The framework

Version Control for Law is a mental model for legal interpretation borrowed from software engineering, specifically from Git. Ross Brodskiy set it out on Medium in April 2026 (Version Control for Law) as a companion to the Gibsonian Canons; the canon names below are those of the Medium edition. The metaphor is not the point. The point is that the rules for changing legal meaning can be written as governance: who may merge, what a commit must contain, which changes are forbidden, and which are allowed only through a logged gate.

The mapping

The force-push gate

Overruling is the most dangerous operation in the model because it changes what the main line of authority says. The essay allows it only when the overruling opinion does all four of the following:

1. Names the violated canon, identifying which interpretive rule the earlier decision broke.
2. Supplies the canonical re-reading, stating the reading the canons actually require.
3. Addresses reliance interests, accounting for those who built on the earlier decision.
4. Reasons textually, resting on the text rather than on changed values, preferences, or politics.

An overruling that fails these conditions is "itself void as precedent." The normalized canons carry the same structure in Stare Decisis Nisi Manifeste Erratum: manifest error under the canons (disagreement with an outcome is not enough), a fair treatment of reliance, and an overruling that is itself reasoned under the canons. Since Supplement S1, a preserved dissent may be invoked as a reason to reconsider, but it cannot displace precedent on its own; any overruling still has to pass the gate.

Style guide versus governance

The essay's sharpest contrast is with Scalia and Garner's Reading Law (2012). A catalog of canons like that works as a style guide: sound advice an interpreter may follow or not. The Gibsonian canons, read through this model, are repository governance: a branching model, a commit policy, a signed-commit requirement, dependency pinning, and protected-branch settings. A style guide improves code. Governance decides what may be merged.

How to apply it

Take the constructed dispute from the Gibsonia manuscript Method Before Power and extend it. Gibsonia's constitution guarantees free movement between inhabited districts (Article 4) and lets the Council restrict movement temporarily when a verified life-support failure creates an immediate danger (Article 12). A statute lets the Administrator close transit corridors during a declared emergency.

Now suppose an early decision upheld a thirty-day closure that continued after the danger ended, reasoning only that an emergency had been declared. It never engaged the words "immediate" and "temporarily." One judge dissented on exactly that ground.

• Is the early decision a signed commit? No. It gave no textual basis for reading the time limits out of Article 12 and never addressed the counter-reading. Under the model it carries reduced weight as precedent.
• The dissent is a preserved branch. It sits in the record. When a new closure is challenged, the dissent is a reason to reconsider, not binding authority.
• Overruling is a force push, so the gate applies. The later court (1) names the violated canons, Verba Suprema and Nulla Superfluitas, because the early reading leaves "immediate" and "temporarily" doing no work; (2) supplies the canonical re-reading, under which closure is lawful only while the verified danger lasts; (3) addresses reliance, for instance by applying the ruling prospectively to operators who organized logistics around the early decision; and (4) reasons from the text, not from a sense that attitudes to emergencies have changed.
• A disallowed dependency is caught. Counsel cites an Earth court's emergency-powers ruling. The court may read it for illumination but cannot rest its holding on it.
• An amended commit is refused. A party argues that "temporarily" was a drafting slip. Unless contemporaneous records show a clerical error with a single corrected reading, the text stands.

The resulting record reads like a clean commit log: what changed, which rule authorized the change, what it cost, and which preserved dissent it drew on.

Known limitations and critiques

Force push is the wrong verb in one respect. In Git a force push can discard history. Legal overruling never erases the earlier opinion; it stays in the reports and remains citable for what it was. The model's own preserved-branch rule compensates, but "force push" should be read as moving the authoritative tip under guard, not as deletion.

Merging law is semantic, not textual. Git detects conflicts by comparing lines. Legal conflicts concern meaning and are found by judgment. The model governs how conflicts are resolved and recorded; it cannot detect them mechanically.

"Signed" is a judgment, not a signature. Anyone can verify a cryptographic signature. Whether an opinion did its textual work is a contested assessment, and "unsigned" could become a label for discounting decisions one dislikes. The four gate conditions narrow that risk without removing it.

Several canons sit outside the normalized text. Versio Enactionis, Relatio Certa, Scriptoris Error, and Manifestatio Ambiguitatis come from the Medium edition. The normalized twenty-three-canon text of Supplement S1 does not contain them, so parts of the mapping currently rest on canons outside the instrument the Institute maintains.

Earth has no version pin. Versio Enactionis presupposes an enacted, versioned interpretive method. Existing legal systems have none, so on Earth the model works as a critique and a design target rather than a description.

The neutrality claim is reported, not established. The essay reports testing the framework against landmark decisions, including Dobbs, Heller, Bostock, Shelby County, and Brown, and finding that it "does not track left or right. It tracks whether the opinion did its textual work." Until the test protocol is published in replicable form, that finding should be treated as a claim. The model also inherits the canons' text-first, liberty-oriented commitments, which are choices rather than neutral defaults.