Baseline-First Audits: Why the First Scan Establishes Truth, Not Scores
Velocity, drift, and decay are comparative by definition. The first audit cannot compute them — and pretending it can produces wrong data. GTL handles this honestly.
Some measurements require at least two data points. Velocity — the rate of change of authority — requires a previous value to compare against. Semantic drift — the shift in topic cluster over time — requires a previous embedding snapshot to compare against. Trust decay, the modeling of freshness loss, requires a previous timestamp to compute against. On the first audit of a domain, none of these can be computed.
The Wrong Approach: Fill With Defaults
A system without GTL might handle this by returning default values: velocity = 0 (neutral), drift = 0 (no drift), decay = 0 (no decay). These are not computed results — they are fabricated placeholders that look like computed results. A user looking at a velocity score of 0 on their first audit might conclude their authority is flat. The correct conclusion is: we do not know yet, and we will know after the second audit.
The GTL Approach: Establish the Baseline, Report It Honestly
GTL handles first audits with state: "empty" and reason: "baseline_established". The temporal authority record is written to the database — it contains all the current-state data needed for future comparisons (embeddings, entity confidence scores, citation scores). But the response declares honestly that velocity, drift, and decay scores are not available yet, and explains why.
●This is not a failure state. It is the system behaving correctly. The first audit establishes the truth baseline against which all future changes are measured. That is a real, valuable output — even if it does not produce a set of scores.
What the UI Shows on First Audit
The SiteNexis temporal authority tab on a first audit shows: "Baseline established on [date]. Velocity, semantic drift, and trust decay analysis activate on your next audit." A progress indicator shows that the first snapshot has been stored. The user understands the system is working, understands what will be available next time, and is not shown fabricated scores. This is state-aware UI honesty in practice.
GTL Systems — Part 7 of 10