Responsible: Lake Watkins
Ecosystem relationships: SingularityNET Foundation, OpenCog Foundation, TrueAGI, EARTHwise Ventures
Status: Active. MAGUS milestones M2–M4 documented as complete in MeTTa. Several Magi Assistant components deployed (GM, Discord, Archive MCP); others code-complete or in development (Smart Glasses, Foundry Phase 2). Inkling game demo in progress.
Magi is an applied AGI initiative developing neuro-symbolic frameworks on top of Hyperon. The project combines the transparency of symbolic reasoning with the adaptability of neural networks, organized around four differentiators: transparency (visible reasoning processes), consistency (reliable outputs), plasticity (genuine learning and adaptation), and corrigibility (meaningful human correction).
Magi's core technical contribution is MAGUS (Modular Adaptive Goal and Utility System), a goal-driven decision-making framework implemented in MeTTa that builds on the OpenPsi/MetaMo heritage while incorporating practical utility-AI patterns from game development. MAGUS introduces the Overgoal — a supreme evaluative mechanism that dynamically promotes and demotes goals based on measurability and cross-goal correlation, preventing instrumental convergence.
Alongside the core framework, Magi develops a suite of AI assistant tools that serve as practical testbeds and products: an AI Game Master for live tabletop RPG sessions, a wearable smart glasses interface for AI coding agents, and the Magi Archive MCP server powering wiki-based knowledge graphs. The ai-server, a shared codebase with EARTHwise, integrates AIRIS, the Elowyn Tree AI chatbot, and the EARTHwise Alignment Benchmark framework.
Magi's 5-year strategic plan progresses from Hyperon integration (Phase 1, current) through social agents, an ownership marketplace, AI game mastering, and ultimately non-gaming applications in healthcare, urban planning, and financial services.
See Magi Full for detailed coverage of the MAGUS framework, tool suite architecture, and implementation status.