Hyperdimensional computing, schema-on-write memory, and deterministic runtimes — published as running code. This is the path from a research idea to a product you can run.
Schema-on-write: every fact mapped into a universal 7-layer × 33-role lattice at write time. Recall becomes a deterministic scan — exact answers, versioned history, no model in the read path. Speaks MCP; SQLite or Postgres.
Ingest, query, and stream, with glyph storage, websocket listeners, and runtime analytics. It hosts the HDC models and MCP tools right next to your agents — state, memory, and compute over open protocols.
An open catalog of pure-HDC models — classification, intent parsing, prompt-injection detection, and more. Character-level encoding, morphology, and syntax, with no LLM in the loop.
OpenRotor — the open runtime for RotorSpec, the standard for deterministic AI agent loops. The machinery under every product: tools, state, and compute, desktop to cloud.
The same governed loop Glyphh Desktop is built on, with Ada memory wired in from the first turn — so the agent remembers the codebase, the decisions, and you.