What the paper argues
Read the formal paper through the hosted PDF. This page is the bridge into the site: it keeps the abstract, concept path, and course path close to the paper instead of sending readers straight into an isolated file.
This paper is the expanded, illustrated path through the Calyx thesis. It starts with the same refusal to flatten that appears in the core course: one input is not one vector. It is a constellation measured by a panel of independent lenses, and the relationships between those lenses are first-class data.
From there, the paper makes a hardware argument. Calyx does a large amount of sparse association work: similarity search, top-k ranking, graph traversal, associative recall, and streaming insertion. Those operations are candidates for an event-driven, memory-near fabric. The paper is careful about the trust boundary: neuromorphic hardware can accelerate precision-tolerant association, but the hash ledger, anchors, and fail-closed decisions remain bit-exact.
The bridge into Concepts
- the record shape — Constellation and Panel;
- the operating grammar — the four verbs and Calculus of Association;
- the trust contract — Grounding & anchors and Fail-closed doctrine;
- why verification holds — Ledger & provenance;
- the refusal to flatten — No-flatten doctrine.
The local visual-edition source and figure manifest are used to author the Stratum E lessons. The paper surface is a research bridge, not a claim that every pre-1.0 mechanism has shipped.