§ 04 · Instrument · Prototype (illustrative)
Freshness Lag — Instrument (Prototype)
Planted sentinel pages with known publish times, probed every 6 hours. The lag between a page existing and a vendor being able to retrieve it.
—median time-to-retrievability · pending first crawlawaiting data
Sentinel pages minted with a unique token and deployed to sentinel.arlenkumar.com; probed for first appearance with a not_indexed vs ranked_below_k discriminator. The Kaplan–Meier time-to-retrievability curve (right-censored 30d) is pending the first crawl — no median hours are reported yet.
time-to-retrievability survival curve · web_search-2026-q3
No data yet. Sentinel pages are deployed and the probe loop is running, but no vendor has been observed indexing a sentinel page long enough to estimate a survival curve. Per-vendor retrievability medians will appear here once the first crawl completes; until then every figure is —, never 0.
§ 04.1
Per Vendor
| Vendor | retrievability h | probe status | censored @30d % |
|---|---|---|---|
| Exa | — | probing · awaiting first index hit | — |
| SerpAPI | — | probing · awaiting first index hit | — |
| Brave | — | probing · awaiting first index hit | — |
| Tavily | — | probing · awaiting first index hit | — |
The instrument is built and deployed; the lag curve is pending the first crawl. Each probe records whether the sentinel token is not_indexed or merely ranked_below_k, so a no-show is distinguished from a low rank. No median hours are published until real observations land — every cell is —, never 0.