Architecting the Neuro-Symbolic Enterprise
The epistemological crisis in financial AI demands a transition from naive "LLM Wrappers" to sovereign, neuro-symbolic architectures. This artifact serves as the physical manifestation of Domain-Driven Design (DDD) and Sly Data telemetry within the Adam repository.
Dual-Process Cognition
Adam is categorized as a Type 3 (Neuro | Symbolic) system under the Kautz Taxonomy. It explicitly separates the probabilistic adam_swarm (System 1) from the deterministic logic of the adam_graph (System 2) to prevent the "Lethal Trifecta" of AI hallucination in capital allocation.
10k-Token Optimization
To facilitate autonomous LLM maintenance without succumbing to "Attention Decay", Adam enforces strict Bounded Contexts. All prompts and JSONLogic are evicted to an assets/ directory, while adam_interfaces governs strict Protocol decoupling.
Repository Architecture Mapping
adam_interfaces/
Python Protocols defining cross-module I/O contracts.
adam_swarm/
System 1: High-velocity unstructured data ingestion.
adam_graph/
System 2: Neuro-symbolic DAG planner and memory.
adam_finance/
Deterministic math (ICAT, SNC, Fama-French). Zero LLM logic.
adam_governance/
Gatekeeper, Proof of Thought logger, and Consensus Engine.
assets/
Evicted prompts and jsonLogic rule matrices.
Immutable Proof of Thought Ledger
This interactive log represents "Sly Data" in action. Operational state is passed deterministically between bounded contexts, generating an auditable trail of how unstructured data was converted into a mathematical risk action.
Deterministic Math Output
Output from adam_finance. Demonstrates the execution of pure, zero-LLM-dependency mathematical logic recalculating risk premiums and limits.
Synthesis Context & MCP Bridge
Demonstrating how an external LLM agent utilizing MCP accesses this exact telemetry bundle to provide reasoned, deterministic decision support without hallucinating internal architectures.
1. Observation
2. Financial Reasoning