VIDEO
HIGHLIGHT
Log InSign up
Hyperclaw: quick thoughts... cognitive orchestration layer for the road to AGI
SummaryTranscriptChat

Hyperclaw: quick thoughts... cognitive orchestration layer for the road to AGI

AI R&D Progress at Singularity ASI Alliance

Introduction to Open Claw

  • Ben Girtzil introduces the rapid progress in AI R&D at Singularity ASI Alliance and his intention to communicate updates through short videos.
  • He discusses "Open Claw," an open-source product that utilizes LLMs (Large Language Models) for automating computer tasks, highlighting its potential security risks.

Security Considerations

  • Girtzil emphasizes the importance of using Open Claw cautiously, suggesting separate devices or virtual machines for experimentation to mitigate security risks.

Impact on Software and AI Industries

  • The advent of Open Claw is seen as a significant win for the open-source community, emerging from independent developers rather than large corporations.
  • He compares OpenAI's approach to launching LLM technology despite known issues (hallucinations), which allowed them to gain a competitive edge over Google.

Automation and AGI Concerns

  • Girtzil notes that while big tech hesitated to fully automate systems due to security concerns, open-source developers are more willing to experiment with innovative solutions.
  • He raises concerns about launching flawed AGI systems but believes that Open Claw represents a positive step forward in automation technologies.

Limitations of Current Technologies

  • Despite its capabilities, Open Claw does not yet represent a breakthrough towards AGI; it lacks true understanding and collaborative research abilities like humans possess.

MetaClaw Development

Introduction of MetaClaw

  • Patrick Hammer from Singularity's AI team developed "MetaClaw," a simplified version of Open Claw written in the novel AGI programming language called Meta.

Features and Capabilities

  • MetaClaw replicates core functions of Open Claw with only a few hundred lines of code, allowing integration with various AI functionalities within an agent framework.

Long-term Memory Integration

  • The development includes structured symbolic long-term memory features, enhancing the capabilities of agents built on this platform.

Self-improving Code Potential

  • The design philosophy behind Meta allows AI systems to revise their own code easily, promoting self-improvement and adaptability.

AGI and Cognitive Process Orchestration

Overview of Metaclaw Integration

  • The speaker discusses a concise, elegantly coded open claw type system that can integrate with AGI and ASI chain tooling, emphasizing its self-improving capabilities compared to the original TypeScript code base.

Development of Quester Product

  • The team is working on a product called Quester, designed as a research assistant for complex projects in science, finance, and business using LLMs and Hyperon tools.

Utilizing Metaclaw in Quester

  • There is potential to use Metaclaw as core infrastructure for the Quester product, allowing it to coordinate various tools effectively.

Cognitive Process Orchestration Challenges

  • The speaker reflects on cognitive process orchestration within AGI systems, referencing the previous cog server from OpenCog that managed reasoning and learning processes.

Introducing Hyperclaw Concept

  • A new concept called Hyperclaw is introduced to run atop Metaclaw, aimed at organizing cognitive processes towards shared goals through an attentional meta protocol.
Video description

Paper: https://drive.google.com/file/d/1WuFnXllMbe5l5f7R2oJLp3JC1JGchqQK/view HyperClaw is a cognitive orchestration layer designed to coordinate heterogeneous AI systems, including LLMs, symbolic reasoners, and neural subsystems. Developed by Ben Goertzel in February 2026, it extends the METTACLAW agentic system into a framework capable of managing multi-module workflows. The architecture consists of three main elements: - Context Frames: Typed structures that represent the current state of a cognitive process, acting as a persistent layer for memory. - Attention-Metaprotocol: METTA rewrite rules that determine which modules are attended to at fast (task execution) and slow (strategic steering) timescales. - Module Spaces: A uniform interface (using HYPERON) that wraps heterogeneous backends, allowing the orchestrator to interact with various AI tools. The system is deployed in two phases: - HyperClaw v1: Focuses on automating complex multi-LLM research workflows, such as quantitative finance, utilizing self-hosted open-source models and commercial APIs. - HyperClaw v2: Integrates into the PRIMUS cognitive architecture, where modules like PLN (uncertain reasoning) and MOSES (evolutionary search) interoperate over a shared ATOMSPACE. The system is designed to run on ASI: CHAIN, featuring capability-secured execution where modules operate under strict, least-privilege contracts. It supports decentralized deployment, multi-party collaboration via DAOs, and economic metering.