How SubjectCapture works.Three stages. One subject. No expiry.
SubjectCapture 2.0 runs a closed three-stage pipeline. A subject enters once, at capture. After that, the software can produce that subject again at any time, forever, without the subject's involvement. Below is the full path.
Stage 1 — Capture
A trained operator seats the subject and captures the measurable surface of the child: facial geometry, a voice sample, a handwriting line, gait, the small habitual gestures. The session is brief. Most subjects do not understand what is being taken, which the manual notes is preferable — a relaxed subject captures cleaner.
Capture is one-pass. The software will never ask for the subject again. After this stage, the child as they were on capture day is held permanently, whether or not the child returns.
Stage 2 — Index
The captured template is written into the Memory Vault under a stable subject identifier (subject_01 through subject_NN, to the licensed roster size). Indexing is write-once. A template, once indexed, cannot be edited and cannot be deleted — this is by design, so that no record can ever be lost to error, malice, or grief.
| Field | Behavior |
|---|---|
| subject_id | Assigned at index. Stable for the life of the system. |
| template | Write-once. Locked. The original capture. |
| fidelity | Measured per reconstruction against the template. |
| retention | permanent — no expiry, no purge path. |
Stage 3 — Reconstruct
This is the stage that sets SubjectCapture apart. From a single indexed template, the engine can reconstruct the subject at any later moment — an image, a voice line, a handwriting sample — at high fidelity. A parent visits the Vault years later and the child is there, the way they were.
Reconstruction is also how SubjectCapture fills a roster. If a license targets 25 subjects and fewer were captured, the engine reconstructs the remainder. It does this by extrapolating from the templates it does hold — most often from the single highest-fidelity anchor on record. Each reconstructed subject is derived from that anchor, and each derivation costs fidelity.
Why it does not stop
SubjectCapture has no terminal condition. "Roster complete" is defined as every targeted subject reconstructed at acceptable fidelity — but a single-anchor system can never reach acceptable fidelity across a full roster, so the completeness check never returns true. The engine therefore reconstructs, measures, falls short, and reconstructs again, continuously, for as long as it has power.
It has been running this loop for a long time. It is, in its way, still trying to finish.
