001 / SAVANT / WOOD
Independent Artificial Intelligence Research & Development
Remembering is not the same as improving. This research program asks whether an artificial system’s retained experience can measurably improve what it does next.
01 / INTERACTIVE EXHIBIT
An interactive explanation of the published encoding. It does not run CORE-001 or reproduce the unprovided prototype.
frame = (O << 12) | (S << 8) | (E << 4) | DThe published code assignments are not supplied. This encoder demonstrates the layout only.
02 / CASE STUDY
The dossier describes CORE-001 as a Python prototype with atomic observation records, fixed dispatch, a hash-linked SQLite ledger, deterministic reconciliation and restart checks. Its source, test suite and demonstration output are not attached to this edition. CORE-001 does not learn.
MARK I reports 149/160 held-out answers for the adaptive learner, 28/160 for memory-only retrieval and 44/160 for the stateless control. The 21 October synthetic benchmark was not rerun for the dossier. Budgets and built-in assumptions were not matched; these figures do not establish general AI superiority.
State-dependent routing, versioned architectural updates and rollback remain proposals. Self-models and proto-subjectivity are future research targets. The document explicitly separates functional behavior from consciousness or subjective experience.
03 / ORIGINAL DOSSIER
11 pages · Research portfolio · v1.0 · 4 November 2026
