Hyperon Whitepaper 2025
Draft — This content has not been approved for publication.
How to Use This Source
Hyperon for AGI⇒ASI: Whitepaper 2025, Deepish-Dive Version
Ben Goertzel, November 19, 2025. ~63 pages, stored across 5 chunks in RawData (chunks 1–4 contain the paper; chunk-5 is a closing sentence only).
This is the primary architectural reference for the current Hyperon stack. It describes the system design, language stack, PRIMUS cognitive architecture, major algorithmic advances since 2023, neural-symbolic integration, AGI→ASI safety planning, and application pilots.
Reading strategy: The whitepaper is dense and layered. For orientation, read §1 (Executive Summary) and §4 (PRIMUS overview). For specific subsystems, use the section map below to jump directly to the relevant chunk. Cross-reference the "Cards Grounded" column to find wiki articles that already synthesize each section.
Section / Chunk Map
| § | Section | Chunk | Cards Grounded |
|---|---|---|---|
| 1 | Executive Summary (1.1 Core Vision, 1.2 Technical Architecture, 1.3 Neural Integration, 1.4 PRIMUS, 1.5 Advances Since 2023, 1.6 From AGI to ASI) | chunk-1 | About Hyperon |
| 2 | Hyperon System Design (2.1 Technical Foundation, 2.2 Space API, 2.3 MORK, 2.4 Neural Integration, 2.5 Distributed Execution, 2.6 State Management) | chunk-1 | MORK Full, DAS Full, Knowledge Substrates |
| 3 | The Language Stack (3.1 Multi-Layer Approach, 3.2 Technical Stack, 3.3 MeTTa, 3.4 MeTTa-IL, 3.5 PeTTa/JeTTa, 3.6 MORKL/MM2, 3.7 PyMeTTa, 3.8 Distribution, 3.9 Dev Workflow) | chunk-2 | MeTTa Full, MeTTa Implementations |
| 4 | PRIMUS Cognitive Architecture (4.1 Goal-directed/Ambient, 4.2 Cooperation, 4.3 Composability, 4.4 Long-time Components, 4.5 New Ingredients, 4.6 Shared Controls) | chunk-2 | PRIMUS Full |
| 5.1 | Weakness Theory (PLN-Quantale Framework, Neural Applications) | chunk-2 | Weakness Theory, PLN Full |
| 5.2 | Geodesic Inference and Control (Mathematical Foundation) | chunk-2 | Geodesic Inference |
| 5.3–5.4 | Fluid Dynamics for Attention / Incompressible-Fluid Networks (ECAN + Optimal Transport, HJB, Helmholtz-Hodge) | chunk-2 | ECAN Full |
| 5.5–5.6 | Integration with Predictive Coding / TransWeave Connections | chunk-2 | No dedicated card — bridging material across PC-Transformers and TransWeave |
| 5.7 | Schrödinger Bridge Learning (Abstract→Detailed models, Predictive Coding, Evolutionary applications) | chunk-2 | Schrödinger Bridge Learning |
| 5.8 | TransWeave (Mathematical Foundations, Impossibility Results, Selective Transfer, Commutativity) | chunk-2 | TransWeave |
| 5.9 | SubRep (CDS/PDS, Motive Geometry, Planner-safe Composition) | chunk-3 | SubRep |
| 5.10 | MetaMo (Pseudo-bimonad, Appraisal/Decision, Operational Design) | chunk-3 | MetaMo Full |
| 5.11 | Algorithmic Chemistry (Reaction Networks, PRIMUS Integration) | chunk-3 | Algorithmic Chemistry |
| 5.12 | WILLIAM-on-MORK (Adaptive Compression, Neural Acceleration) | chunk-3 | WILLIAM |
| 5.13–5.14 | How Advances Work Together / Path Forward | chunk-3 | PRIMUS Full (partial) |
| 6 | Re-Visioning Longstanding Components (6.1 PLN, 6.2 MOSES/GEO-EVO, 6.3 ECAN, 6.4 Pattern Mining, 6.5 Concept Blending) | chunk-3 | PLN Full, MOSES Full, ECAN Full, Reasoning and Search |
| 7 | Neural-Symbolic Synergy (7.1 Two Paths, 7.2 Symbolic Heads, 7.3 Pattern Mining as Prior Discovery, 7.4 QuantiMORK, 7.5 Weakness-Based Stability, 7.6 WILLIAM, 7.7 Workflows, 7.8 Deeper Significance) | chunk-3 | Neural-Symbolic Integration, Semantic Parsing Full, QuantiMork |
| 8 | AGI→ASI Transition (8.1 Self-Improvement, 8.2 Goal Stability, 8.3 Global Regulators, 8.4 Self-Modification Pipeline, 8.5 Safety, 8.6 Governance, 8.7 Upgrade Example, 8.8 Checklist) | chunk-4 | Self-Modification and Safety |
| 9 | Application Domains (9.1 Game AI, 9.2 Social Robotics, 9.3 Bioinformatics, 9.4 Mathematics) | chunk-4 | Applications (overview), Game AI, Social Robotics, Bioinformatics, Mathematics, Game Worlds, Robotics, Bio-AI |
| 10 | Research Agenda (Efficiency, Stability, Transfer, Benefit metrics) | chunk-4 | No dedicated card |
Key Concepts by Section
| Concept | Primary Section | Also Appears In |
|---|---|---|
| Space API | §2.2 | §3.3, §3.8 |
| MORK / PathMap / Triemap | §2.3 | §3.6, §5.3, §7.4 |
| ShardZipper | §2.3.2 | §5.3.2 |
| ByteFlow / PPTNs | §2.3.2 | §3.6, §7.4 |
| QuantiMORK | §2.4.1 | §7.4 |
| MeTTa language / MeTTa-4 | §3.3 | §3.4 |
| MeTTa-IL / GSLT | §3.4 | §3.2 |
| PeTTa / JeTTa | §3.5 | — |
| MM2 / MORKL | §3.6 | §6.1 |
| PyMeTTa / metta-magic | §3.7 | — |
| Weakness Theory / Quantale | §5.1 | §6.1, §7.5 |
| Geodesic Inference | §5.2 | §5.8.3, §6.2 |
| ECAN / Fluid dynamics | §5.3–5.4 | §6.3 |
| Schrödinger Bridge | §5.7 | §5.7.2 (MOSES) |
| TransWeave | §5.8 | §5.6, §7.5 |
| SubRep (CDS/PDS) | §5.9 | §5.10 |
| MetaMo (pseudo-bimonad) | §5.10 | §5.9, §7.5.3 |
| Algorithmic Chemistry | §5.11 | — |
| WILLIAM compression | §5.12 | §7.6 |
| PLN / Factor graphs | §6.1 | §3.6.1, §5.1 |
| GEO-EVO (MOSES) | §6.2 | §5.7.2 |
| Symbolic Heads | §7.2 | — |
| Goal stability / Lyapunov | §8.2 | §8.4, §8.5 |
| Self-modification pipeline | §8.4 | §8.7 (worked example) |
| RSpace / Rholang / ASI Chain | §2.5, §8.6 | §3.8 |
Enrichment Targets
Most major whitepaper sections now have at least some wiki coverage. The following are candidates for deeper technical enrichment during a second pass, where the existing card could benefit from additional whitepaper-sourced detail:
| Section | Existing Card | Priority | Enrichment Opportunity |
|---|---|---|---|
| §5.5–5.6 | None — bridging material | Medium | Cross-cutting integration between PC-Transformers and TransWeave; could become a section in PRIMUS Full or a standalone bridging card |
| §5.8 | TransWeave | High | Mathematical foundations (impossibility results, selective transfer, commutativity) could be expanded with a +full companion |
| §5.12 | WILLIAM | High | Adaptive compression details and neural acceleration specifics could be expanded with a +full companion |
| §8 | Self-Modification and Safety | High | Self-modification pipeline (5 stages), goal stability math (Lyapunov), and decentralized governance details are dense enough for a +full companion |
| §6.5 | None | Low | Concept blending — brief treatment in whitepaper, may not warrant standalone card |
| §10 | None | Low | Research agenda metrics — useful but less urgent than architectural content |
Best Source For...
- Understanding the overall architecture: §1 (chunk-1) + §2 (chunk-1)
- How MeTTa compiles and executes: §3 (chunk-2), especially §3.3–3.6
- PRIMUS cognitive cycle: §4 (chunk-2) for overview, §5 (chunks 2–3) for specific mechanisms
- Math behind any PRIMUS component: §5.x (chunks 2–3) — each subsection has its own "Mathematical Foundation"
- How PLN works in the new stack: §6.1 (chunk-3) for quantale factor graphs, §5.1 (chunk-2) for weakness theory
- Neural-symbolic integration approaches: §7 (chunk-3), especially §7.2 (Symbolic Heads) and §7.4 (QuantiMORK)
- Safety and self-modification: §8 (chunk-4) — the only detailed treatment
- Application domain requirements: §9 (chunk-4) — Minecraft, robotics, bioinformatics, mathematics pilots
- MORK internals (ShardZipper, ByteFlow): §2.3 (chunk-1)
- ASI Chain / decentralized execution: §2.5 (chunk-1) + §8.6 (chunk-4)
Tags
Discussion