First preserved complete-window ASG run.
A requested window of ten live Google-via-Serper candidates was fully established, frozen in provider order, evaluated by TA-14, and accounted for at the delivery boundary. This is a complete-window benchmark, not a claim that every ALLOW determination is objectively or permanently correct.
Ten established. Ten evaluated. Ten delivered.
The admitted evidence record preserves original provider rank, delivered rank, title, URL, snippet, host, determination, reason code, explanation, and evaluation tier for all ten candidates. The companion non-admitted record explicitly records a count of zero and an empty non-admitted set. Nothing is inferred from absence.
10 preserved entries
Every delivered candidate remains bound to its original provider rank and its TA-14 determination evidence.
0 preserved entries
The zero is explicit evidence: no established candidate received HOLD, DENY, or ESCALATE in this run.
What this benchmark establishes — and what it does not.
It demonstrates a live bounded provider capture, preservation of original rank, independent TA-14 determination, delivery accounting, and paired admitted/non-admitted evidence. It does not prove universal search truth, does not certify that every determination is correct, and does not establish governance over Google's internal computation. A later changed context may require a new determination.
Two benchmarks. Two different things proven.
The mixed-delivery record demonstrates that established candidates can be stopped at the delivery boundary and remain preserved. This complete-window record demonstrates full requested-window establishment and end-to-end accounting when every candidate receives ALLOW.
View mixed-delivery benchmark →