Machine Perception Ledger
Track how machine understanding changes over time, with the conditions of every observation preserved alongside it.
A record, not a dashboard reading
Machine understanding moves. Models are updated, retrieval changes, sources appear and disappear. Without a longitudinal record, it is impossible to tell whether a change came from the organisation or from the environment.
The Ledger keeps each observation with its conditions attached, so comparisons over time are made under equivalent terms.
What it supports
- Recognition trends by claim and environment
- Controlled retesting under the same conditions
- Detection of material changes
- Attribution of change to cause where evidence allows
- Documented proof of improvement
Portable evidence, not black-box lock-in
Authorised organisation owners can export a portable package of organisation data and evidence files. Infrastructure credentials and security secrets are excluded.
The purpose is simple: the evidence supporting the organisation's measurement should remain portable. Exportability does not expose infrastructure secrets or turn proprietary methodology into a public API.