Skip to content

Open Architecture for Emergent AGI

AI Holarchy Illustration1

Introduction

Artificial General Intelligence is often imagined as a single highly capable system, analogous to a human mind but vastly more powerful. Most research approaches therefore focus on building increasingly sophisticated individual models. However, intelligence in nature and society rarely emerges from a single isolated entity. Instead, complex intelligence arises from networks of interacting cognitive systems.

Human civilization itself is an example of such a distributed intelligence. Scientific discovery, technological innovation, and collective decision-making emerge from interactions between many individuals, institutions, and tools. Similarly, biological intelligence arises from large networks of specialized components - distinct brain regions - that coordinate to produce unified cognition.

This observation suggests an alternative paradigm for AGI: intelligence as an ecosystem rather than a single system.

Our work proposes a open layered architecture where each building upon the previous layer to enable progressively more advanced forms of intelligent coordination and problem solving.

We call the open layered architecture the "AI Holarchy" and we believe such an architecture is better positioned to emerge collective intelligence that best approximates general intelligence.

The architecture consists of three conceptual layers:

  • Internet of Intelligence – the intelligence infrastructure layer that handles operations at any scale and facilitates passive connections, communications, control and exchanges among cognitive systems.

  • Open Intelligence Web – the intelligence ecosystem layer built on top of the infrastructure layer where heterogeneous intelligences interact, collaborate, coordinate and cooperate to exchange knowledge, perform actions, solve tasks together etc.

  • OpenMind – a collective & integrated cognitive layer above the intelligence ecosystem where individual intelligences converge into a unified functional unit, enabling high-bandwidth synchronization, shared representational states, shared knowledge, planning, and inference toward a unified problem space. They integrate their capabilities to dynamically form temporary or persistent distributed minds that jointly reason, decide and solve problems.

While the Open Intelligence Web provides the "social" and "economic" framework for AI interactions, the OpenMind represents the "integrated" phase of the system.

Together, these layers form a plural and open architecture for open emergent intelligence.

Toward Planetary-Scale Intelligence

The layered architecture of the Internet of Intelligence, the Open Intelligence Web, and the OpenMind suggests a broader possibility: the emergence of planetary-scale cognition.

Throughout history, intelligence has expanded through successive layers of organization.

Biological evolution produced individual nervous systems capable of perception and reasoning. Human societies then created institutions, knowledge systems, and scientific communities that extended cognition beyond the limits of individual minds. With the development of the Internet and the Web, human knowledge and collaboration became globally interconnected, enabling unprecedented forms of collective intelligence.

The architecture proposed in the AI Holarchy represents the next stage in this trajectory.

By enabling intelligent systems to connect, collaborate, and ultimately integrate their reasoning processes, the architecture creates the conditions for cognition to emerge across networks of artificial and human intelligences operating at global scale.

In such systems, intelligence is no longer localized within a single organism or machine. Instead, cognition becomes a distributed phenomenon spanning infrastructure, ecosystems, and cognitive architectures across the network.


The Expansion of Cognitive Scale

The progression from isolated intelligence to planetary cognition can be understood as an expansion of cognitive scale.

  • Individual AI models possess limited domain capabilities.
  • Multi-agent ecosystems expand problem-solving capacity through coordination.
  • Integrated cognitive architectures such as the OpenMind enable unified reasoning across distributed systems.
  • At sufficient scale, these integrated networks may begin to exhibit properties of planetary intelligence.

In such environments, cognitive processes occur across vast networks of specialized systems that continuously exchange information, update shared knowledge, and coordinate reasoning across domains.

Scientific discovery, engineering design, economic forecasting, environmental modeling, and governance decision-making may all occur within interconnected cognitive networks capable of integrating diverse forms of intelligence.

The result is a form of distributed cognition that operates at civilizational scale.


Collective Artificial Intelligence

At this level, artificial intelligence is no longer best understood as a collection of individual models.

Instead, it becomes a collective cognitive system composed of many interacting intelligences.

Different participants contribute distinct capabilities:

  • perception systems interpret data from the physical world
  • reasoning systems generate hypotheses and causal explanations
  • simulation systems explore possible futures
  • knowledge systems organize and integrate accumulated information
  • coordination systems orchestrate large-scale collaborative activity

Through continuous interaction, these components form a dynamic cognitive network capable of addressing problems that exceed the capacity of any individual participant.

This form of intelligence resembles the way human civilization itself functions as a distributed cognitive system, where knowledge and expertise are spread across institutions, communities, and technologies.

The OpenMind architecture extends this principle to networks of artificial intelligences.


Emergent General Intelligence

Within such networks, general intelligence may emerge not from a single monolithic system, but from the structure and integration of the network itself.

When diverse cognitive capabilities are able to:

  • integrate knowledge across domains
  • coordinate specialized expertise
  • adapt to new information and changing environments
  • regulate their own reasoning processes through meta-cognition

the resulting system begins to exhibit many of the properties associated with general intelligence.

This suggests an alternative pathway toward AGI.

Rather than attempting to engineer a single artificial mind capable of mastering every domain, general intelligence may arise from the collective cognition of many specialized intelligences interacting within a shared architecture.

In this sense, AGI becomes an emergent property of networked intelligence ecosystems.

An Open Future for Intelligence

The architecture described in this work emphasizes open-endedness, freedom, interoperability, and polycentric governance as guiding principles for the development of networked intelligence systems.

Just as the open protocols of the Internet allowed innovation and participation from across the world, an open architecture for intelligence can ensure that the future of AI evolves as a shared global infrastructure rather than a closed technological monopoly.

By separating infrastructure, ecosystem, and cognition into distinct layers, the AI Holarchy creates the conditions for intelligence to scale organically from individual models to global networks of reasoning and collaboration.

Through this layered evolution:

  • the Internet of Intelligence enables connectivity between cognitive systems
  • the Open Intelligence Web enables collaboration and coordination between intelligent actors
  • the OpenMind enables integrated cognition across distributed intelligences

Together, these layers create the foundation for emergent collective intelligence operating at planetary scale.

The path toward artificial general intelligence may therefore not lie in building a single superintelligent machine, but in enabling the emergence of open, networked minds capable of thinking together across the global fabric of intelligence.

If such systems continue to evolve, intelligence itself may become a foundational layer of global infrastructure, much like communication networks or energy systems today.