In effect, Hyperon provides the raw “Lego bricks” of AGI; PRIMUS is the architectural recipe that configures and orchestrates them into a unified AGI engine — fully autonomous, self-evolving, and characterized by emergent cognitive synergy. PRIMUS is a meta-architecture specification implemented in MeTTa: a high-level orchestration layer and accompanying configuration library that defines how Hyperon’s modular engines fit together into a cohesive AGI system:
-
Module Topology
Specifies which Hyperon components to invoke, in what order, and how data flows between them (e.g. piping MOSES proposals into PLN inference, feeding ECAN attention cues into NARS).
-
Goal & Motivation Loops
Templates for curiosity-driven search, goal decomposition, reward signals, and learning triggers that animate continuous, self-directed cognition.
-
Attention & Resource Policies
Prescriptive rules for ECAN/ActPC to allocate CPU, memory, and inference budget across competing kernels for maximal efficiency.
-
Integration Contracts
Standardized MeTTa interfaces and API bindings ensuring each kernel—symbolic, probabilistic, evolutionary, neural—can be hot-swapped or scaled independently.
-
Cognitive Synergy Patterns
Reusable coordination motifs (e.g. “evolution→inference→attention” cycles) that underlie emergent generalization and robust decision-making.