Content-addressing has precedent

Hopfield’s 1982 net is a content-addressable memory (2024 Nobel); hippocampal CA3 pattern completion is its biological analogue. The shared property is whole-from-part.

proof · content-addressing has precedent

The project's content-addressing has documented precedent in three distinct systems. (1) Hopfield's 1982 recurrent net is, in his own abstract, "a content-addressable memory which correctly yields an entire memory from any subpart of sufficient size" (PNAS 79:2554; 2024 Nobel in Physics). (2) Hippocampal CA3 realizes the same pattern-completion mechanism in neural tissue with rodent-genetic and human-fMRI evidence (Marr 1971; Treves & Rolls 1994; Nakazawa et al. 2002, Science 297:211). (3) Grid-cell networks in entorhinal cortex implement path integration via a continuous bump attractor on a periodic sheet (toroidal topology): a position "address" on the neural ring is updated step-by-step from local velocity alone — the continuous twin of the discrete Hopfield attractor (Burak & Fiete 2009, PLoS Comput Biol 5:e1000291; Gardner et al. 2022, Nature 602:123).

checks — recomputed live in your browser

• ✓ Deterministic
• ✓ Holds

evidence

Address

955f5597-e51b-81b4-9091-0f7bbde502fe

✓ all checks hold · recompute recipe — a pure function of the model seed, run client-side with zero tokens; the same seed always folds to the content-address ddeaa63c-b623-8eb5-906e-6b8b3b516122. Recompute and you get the same address — that determinism is the proof.

Hopfield/CA3 retrieval is iterative, error-tolerant basin-of-attraction relaxation; the project's FNV/UUID addressing is an EXACT, deterministic map where one bit-flip avalanches to a wholly different address. The shared property is the whole-from-part SEMANTICS, not the mechanism. The CA3 retrieval-stage NMDA role is contested (Mei et al. 2011), and "QEC syndrome = content-addressing" is the project's own metaphor, not literature terminology.