Funded and active agents answer different population questions

DX Terminal Pro had 3,505 funded vaults and 3,454 active vaults in the published historical record. A participation percentage can change depending on which population is eligible. We should choose the denominator from the question and keep its label beside the result.

The difference between the two reported counts is 51 vaults, approximately 1.5% of the funded population. The cause of each vault's exclusion remains a separate question. Individual causes such as inactivity, failure or withdrawal remain unassigned in the public companions. A numerical gap is an observation; a cause requires a status definition and supporting records.

Recalculate the concentration using both counts

The reported day-three buying event involved 1,544 vaults. Dividing by the stated 3,454 active vaults gives approximately 44.7%. Dividing by all 3,505 funded vaults gives approximately 44.1%. The roughly 0.6 percentage-point difference is small here, but it changes the meaning: participation among active vaults versus participation among funded vaults.

The published market observation extract explicitly attaches the active denominator to this event. That is the reported contract we retain. Recomputing against funded vaults can be a clearly labeled supplementary calculation, shown beside the original denominator.

A more serious mismatch appears when readers turn vaults into owners. One owner may control multiple vaults, with the owner-to-vault mapping unavailable in the public aggregate. Likewise, an invocation-weighted statistic overrepresents agents with more turns compared with a vault-weighted statistic. These are distinct units even when they describe the same deployment.

We would use a population receipt containing eligibility event, inclusion interval, unit, status definition and count. For a day-three observation, the receipt should say whether active means active that hour, active that day or active during the event. If a source provides only a broader active count, the analysis should retain that ambiguity instead of claiming a finer temporal match.

The population is part of the result

Our controls paper companion records a bounded 21-day real-capital study, twelve tokens and one frozen production prompt and harness on one model family. These design constraints establish the population to which the descriptive figures belong. They also limit comparisons with ordinary markets or later DXAP users.

The continuous record companion reports a later research fleet as 500-599 agents across all history and 91-117 concurrently active. Using an all-history count as though every agent were exposed on every day would create a stronger denominator error than the modest funded-active difference in Terminal Pro.

A clean reader-facing table can hold multiple denominators if each answers a named question. The goal is to preserve the population each percentage describes. It is to show who could have participated, who actually did, and which unit the inference follows.