CirIAco
Machine-Native Knowledge Architecture & Cryptographic DAG Engine
What is CirIAco?
Current knowledge bases (PDFs, Markdown wikis, document stores) assume human readers: they are linear, narrative-driven, and lack cryptographically verifiable provenance. When Large Language Models ingest these documents, they frequently suffer from citation drift, context pollution, and hallucinated attributions.
CirIAco rethinks knowledge representation specifically for AI ingestion. It decomposes complex domains into discrete, immutable propositions termed Atmo blocks, structured as a Directed Acyclic Graph (DAG).
Core Architectural Pillars
- Machine-Native Atomicity: Information is stored not in pages, but in self-contained, typed proposition nodes (
Atmo). - Cryptographic Provenance: Every proposition is hashed via SHA-256 and linked to its causal predecessor nodes, creating an immutable lineage tree.
- Deterministic Traversal: An LLM does not need to parse noisy conversational prose; it traverses deterministic subgraphs, guaranteeing that every generated token is bounded by verified ground-truth nodes.
- Attribution & Royalty Lineage: Recursively computes informational dependencies, making it possible to trace multi-hop reasoning back to foundational authors.
Integration with LLMs & Agents
CirIAco acts as the verifiable external memory layer for frontier agents. Instead of passing massive unstructured context windows, agents query subgraphs matching specific dependency criteria, ensuring sub-500ms retrieval and zero hallucinated citations.