A private-agent company cannot honestly use conventional behavioral telemetry as its north-star instrument. The open problem is to estimate useful daily agents while treating the operator as an adversary and making individual behavior structurally difficult to observe.
Estimate useful-agent reach from receipt-derived, edge-local sketches with calibrated noise and secure aggregation, without reconstructing a person’s behavior or a trajectory over time.
The usual growth dashboard measures the part of a product that phones home. A local-first agent needs a metric that does not punish the architecture for respecting its own boundary.
This is explicitly an open research problem and a proposed construction. It is not an implemented analytics system.
Primary source: Counting Without Watching
Implement an auditable toy simulation: receipt predicate → local randomized response → time-decoupled batching → secure sum → population estimate with published error bounds.
The companion essay is written for a broader technical audience. The repository and developer community are the places to turn a claim into a contribution.
One is made by Hushh Technologies Corporation: private, sovereign AI that runs on what you own.
One is made by Hushh Technologies Corporation: private, sovereign AI that runs on what you own.