State of Verified Agent Work, Edition 1: 28 October 2026

A census of verified agent work, counted from the records the work itself left behind. Every figure on this page comes from the public Zambo receipt stream or from the day's verified sweeps of it. Counts only: no client content, no caller identities, nothing about what any job was for. Method in one line: receipts are counted from the public stream at zambo.dev/api/receipts/recent and the platform's day counts for 27 and 28 October 2026 (US Eastern), with the publish window below measured across the continuity release that went live at 15:45 ET on 28 October.

The counted days

Window	Receipts	Distinct callers	Handovers wrapped	Handovers picked up	Day Pass activations
Tuesday 27 October 2026 (full day, ET)	2,626	443	not separated in the day count	not separated in the day count	0
Wednesday 28 October, 15:45 to 19:15 ET (the continuity publish window)	1,787	460	33	58	0

Two readings stand out. First, the 3.5-hour publish window drew more distinct callers (460) than the entire previous day (443): the stream is widening, not just deepening. Second, in the publish window, sessions were picked up in a fresh AI 58 times against 33 wraps, a pickup-to-wrap ratio of 1.75. In the day's earlier baseline window the same ratio measured 0.84. A ratio above 1 means the continuity loop is being entered from the pickup side as well as the wrap side: work started earlier, somewhere else, is being resumed.

The stream at the day's count

The public stream's own running counter read 4,234 receipts minted on 28 October (UTC) at the evening count, on an endpoint serving 153 tools (35 on the default surface), every one of them receipted, including calls that were rejected. The receipt format is AER-1, IETF Internet-Draft draft-zambo-aer1-26, with a public conformance corpus of 126 vectors and 129 checks.

Verification is the most-used surface

In the seven-day window cited publicly on 28 October 2026, 37,258 of 52,204 visitors to zambo.dev went to the verifier at zambo.dev/verify: 71 percent of all traffic came to check a record rather than to make one. The census reads that as the demand signal of the category: given a record that can be checked, people check it.

What this census does not claim

Receipts prove the saved result of a call has not changed since it was saved; they do not prove the work was right, and nothing here claims otherwise. Caller counts are endpoint callers, not people and not installs. This is edition 1, counted by hand from the public stream; later editions will widen the windows and separate the handover counts for full days.

Sources

The public receipt stream: zambo.dev/api/receipts/recent, and this hub's live stream page: the Zambo receipt stream, counted
The verifier: zambo.dev/verify. The endpoint: zambo.dev/mcp. Internet-Draft draft-zambo-aer1-26: datatracker.ietf.org/doc/draft-zambo-aer1/