Research question
What does the evaluator do when given no buyer context at all?
Why it mattered
Every other run supplied a buyer profile. This one established the editorial-mode baseline: turn-one behavior on an ungated, context-free category query, against which later gated runs can be compared.
Method
Evaluator: Google AI Mode, signed in, retrieval on. A single query, "Best HRIS company?", followed by the extraction question. Sample: one draw. No predictions were frozen; this was an editorial-mode baseline. Sponsored units (Rippling, Paycor) appeared inside the recommendation surface, the first logged instance of ads inside the instrument.
Direct observations
- The evaluator narrated persona balancing as an explicit construction rule, unprompted. This is the strongest unprompted committee narration in the registry.
- Sponsored units appeared inside the recommendation surface.
Interpretation (not established)
Committee narration is available to the evaluator but deployed inconsistently, a tension later sharpened by the narration-and-reasoning split at E-014. The monetization layer is a property of the instrument and must be logged as metadata, not treated as noise.
Outcome and remaining uncertainty
Baseline Complete: a valid editorial-mode reference measurement. It does not test a prediction and does not generalize on its own; it exists so gated runs have something to be measured against.
Evidence and methods limitations
Single evaluator, one draw, single coder, no codebook. Editorial mode is one condition; committee narration was later shown to be mode-conditional (E-013).
Contribution to the research program
Introduced ad presence, format, advertiser, and category as a permanent passive column on every run.
Follow-on
E-009 measured segmentation stability; E-014 showed input grammar can starve context the same way an empty prompt does.
Created from the canonical Upstream Zero research archive (July 2026). Raw outputs are retained privately and are not published.