Knowledge drift is silent by default.

Each mature training library drifts. It is not a matter of author negligence. It is a natural consequence of authoring content, over years, against sources that themselves change and against operational practice that itself changes. The failure is not that drift occurs. The failure is that most libraries have no mechanism to see it occurring.

When drift is finally noticed, it is usually by someone external. An auditor, an incident review, a regulator's question the library cannot answer. By then the divergence has been feeding operational decisions for months or years. The corrective cost is not the cost of updating a module. It is the cost of reconstructing a defensible position after the fact.

Drift is a condition, not an event

Libraries do not fail suddenly. They accumulate small divergences from the sources and the practice they once tracked. Each divergence, on its own, is negligible. In aggregate, over time, they produce a library that no longer says what the organization would want it to say, without anyone having authored the change.

Three forms drift takes

Not one phenomenon. Three. Each has a different source, a different pace, and a different signature. All three are invisible in a library organized as content.

• Source drift. The regulation, standard, or policy changes. If the library has no link back to the source, the drift is invisible until a human notices, which is usually late.
• Interpretation drift. Different authors, at different times, interpret the same source clause differently. Two modules end up teaching subtly conflicting things. Learners see the conflict. Authors do not.
• Practice drift. Operational practice quietly diverges from documented procedure. Trainers, drawing on practice, teach what people do, not what the procedure says. The gap widens with each cohort.

Why libraries organized as content cannot see it

Structure governs detection. All three forms of drift are invisible in a library organized as content, because there is nothing to compare a module against. The source it once satisfied has no explicit link to it. A human review can find drift, but only by reading the library end to end against the sources end to end, and only if the reviewer is competent to spot the divergences. This kind of review is rare, expensive, and by the time it happens the drift has usually already produced an operational consequence somewhere upstream.

From detection to adjudication

Design that begins with the framework changes the detection problem. Each framework node is tied to the source clause that implies it. Each content asset is tied to the node it satisfies. When the source changes, the change is a diff against the framework, not a diff against a document nobody has time to reread. Nodes that require review light up. Content assets tied to those nodes are flagged. Drift stops being something a human has to notice. It becomes something the system reports, and something a human has to decide about. That inversion, from detection to adjudication, is the point.