Demo · synthetic sample data. This is the honest-interface surface only — every node shows a calibrated posterior, an expandable proof, and an explicit “evidence exhausted” terminal. It is not real inference and makes no accuracy claim (that is gated by a real-data benchmark). Source: github.com/andrewhoyle/PastSee.

PastSee — belief graph

✓ confirmed · ~ probable · ? hypothesis · ∅ evidence exhausted

✓ confirmed (0.93) Thomas Ashford (rec-12) ↔ Thomas Ashford (rec-48)

why?

Posterior: 0.93
Evidence: matching surname (Soundex A261); same parish (Ely); birth years agree (1842/1843)
Provenance:
  - field 'surname' of record rec-12
  - pixels (0.10, 0.21) on img://ely-1843-p4

~ probable (0.61) Mary Ashford (rec-12) ↔ Sarah Ashford (rec-31)

why?

Posterior: 0.61
Evidence: plausible mother-child age gap (24y); same place
Provenance:
  - field 'place' of record rec-31

? hypothesis (0.38) John Ashford (rec-12) ↔ John Ashford (rec-77)

why?

Posterior: 0.38
Evidence: shared given+surname only; different county; weak
Provenance:
  - field 'given_name' of record rec-77

∅ evidence exhausted evidence exhausted