Version power
Version power designates an entity’s capacity to make a given canonical version (definition, policy, framework, state) prevail in AI systems, and to make previous versions explicitly obsolete, traceable, and not activatable by default.
In an interpreted web, “updating a page” is not enough. Without version discipline, inertia, trail, and remanence keep previous versions in circulation. Version power is therefore a lever for interpretive sustainability.
Definition
Version power is the set of mechanisms allowing to:
- declare a canonical version (date, status, perimeter);
- make traceable previous versions (history, changes);
- impose prevalence of the current version in activation (links, graphs, surfaces);
- reduce activation of obsolete versions (deprecation, redirections, governed negations).
Version power transforms a canon into a living system, where truth can evolve without losing its interpretive stability.
Why this is critical in AI systems
- AI reuses traces: an old version can persist via secondary sources.
- The web does not forget: aggregators, citations, caches, reproductions maintain the old state.
- Correction is costly: without versioning, each correction becomes a fight against inertia.
Practical indicators (weak version power symptoms)
- An old definition regularly reappears despite a corrected page.
- Two versions coexist without hierarchy (trail).
- AI systems do not know which version is the most recent.
- External sources continue citing the old version as reference.
What version power is not
- It is not merely a number (v1, v2). It is a prevalence and deprecation device.
- It is not a technical obsession. It is a condition for interpretive stability.
- It is not a decorative changelog. It is an obsolescence governance.
Minimum rule (enforceable formulation)
Rule VP-1: any canon subject to evolution must declare a prevailing version and an obsolescence mechanism for prior versions. Without version power, governance cannot guarantee reduction of inertia, trail, and remanence.
Example
Case: a policy is updated, but AI systems continue citing an old page reproduced by an aggregator.
Diagnosis: weak version power (no enforceable prevalence, no activatable obsolescence).
Expected correction: declare the canonical version, reinforce activation of the current version, and neutralize activation paths of the old one (deprecation, redirections, evidence).