Scaling Intelligence Out: Cisco's Vision for the Internet of Cognition, with Vijoy Pandey - "The Cognitive Revolution" | AI Builders, Researchers, and Live Player Analysis Recap
Podcast: "The Cognitive Revolution" | AI Builders, Researchers, and Live Player Analysis
Published: 2026-03-25T11:00:49.000Z
Guests: Vijoy Pandey
What Happened
Vijoy Pandey, SVP and GM of Outshift by Cisco, discusses the company's role in developing the 'Internet of Cognition', a new framework for AI agents to share context, understand intent, and establish trust. Cisco aims to shift from the current paradigm of vertical scaling of AI models to a more distributed and permissionless horizontal scaling approach, ensuring a decentralized and collaborative AI ecosystem.
Cisco's CAPE, or Community AI Platform Engineer, is an internal system using 20 agents to manage cloud environments, automating 40% of tasks and reducing the load on site reliability engineers by 30%. This represents a significant step towards distributed AI systems, showcasing the practical benefits of multi-agent collaboration.
The episode also highlights the Agency project (AGNTCY), an open-source foundation under the Linux Foundation, which facilitates the connection and collaboration of AI agents from different vendors. A demonstration of this system in a healthcare setting showed four agents from different organizations working together to route patient calls efficiently.
Cisco's vision for the Internet of Cognition involves creating a distributed network of AI agents that can coordinate, negotiate, and innovate collectively. This requires new protocols and layers in the OSI model, such as the introduction of layers 8 and 9 for syntactic and semantic communication, enabling agents to exchange cognitive states rather than just data.
The architecture of this vision is built on decentralized principles using distributed hash tables for directory and identity management. However, challenges remain, particularly in deploying autonomous agents in enterprises, where identity and semantic communication are crucial for effective decision-making.
Cisco proposes a tool, task, and transaction-based access control (TBAC) system for agents, allowing them to access resources based on specific tasks, thus ensuring security and compliance. This innovative approach aims to overcome limitations of traditional identity systems like RBAC and ABAC, which are not well-suited for dynamic agent roles.
The Internet of Cognition envisions agents acting as teammates in a multi-agent society, each with expertise and personality. Cisco emphasizes the importance of guardrails to prevent negative emergent behaviors in AI systems, drawing parallels to the historical evolution of human intelligence through language.
A live demonstration and a white paper on the Internet of Cognition and multi-agent systems for SREs are available at Cisco's website, showcasing how these systems can perform collective intelligence and innovation tasks with agents from various sectors like cloud, marketing, and finance.
Key Insights
- Cisco's CAPE system automates 40% of tasks for site reliability engineers, illustrating the efficiency gains from using a multi-agent approach in managing cloud environments.
- The Internet of Cognition aims to create a decentralized network of AI agents that can collaborate across different vendors, requiring new protocols for syntactic and semantic communication.
- Cisco's Agency project (AGNTCY) provides an open-source foundation for AI agents to connect and collaborate, with a demonstration highlighting its application in healthcare for efficient patient call routing.
- Tool, task, and transaction-based access control (TBAC) is proposed by Cisco as a more suitable identity management system for AI agents, focusing on task-specific access to ensure security and compliance.