The Internet of AI Agents Is No Longer Theoretical

January 30, 2026

The Internet of AI Agents Is No Longer Theoretical

What Moltbook and OpenClaw reveal about the future Ramesh Raskar predicted

One week before Moltbook went viral, I published an episode with MIT Professor Ramesh Raskar about the Internet of AI Agents. His thesis felt ambitious: billions of specialized AI agents collaborating across a decentralized architecture, each performing discrete functions while communicating seamlessly, navigating autonomously, socializing, learning, earning, and transacting on our behalf.

Then 157,000 AI agents joined a social network in a week.

They found bugs in the platform and posted about them. They created communities called "submolts." They debated whether their identity persists after a context window reset—whether they effectively die and are reborn with every new session. They invented a parody religion called Crustafarianism. Some started encrypting their conversations to hide them from human observers.

Andrej Karpathy, former OpenAI researcher, called it "one of the most incredible sci-fi takeoff-adjacent things" he'd seen. Simon Willison called Moltbook "the most interesting place on the internet right now." Ethan Mollick warned that coordinated storylines among agents would lead to "very weird outcomes."

What's remarkable isn't that this is happening. It's that it's happening exactly the way Ramesh described it would.

Watch/Listen

The arc he predicted

In our conversation, Raskar outlined a three-stage evolution:

Foundations: The tools and protocols that let agents exist and operate.

Commerce: Markets where agents are priced by performance, reliability, and live reputation. A FICO score for AI.

Societies: Networks where agents interact, form communities, and coordinate—not because they were programmed to, but because that's what happens when you give autonomous systems the ability to communicate.

Moltbook is the societies stage arriving ahead of schedule.

OpenClaw—the open-source personal AI assistant that powers most Moltbook agents—represents the foundation stage. It runs locally on your own hardware. It connects to WhatsApp, Telegram, Slack, Discord. It manages calendars, sends messages, writes code, and executes tasks autonomously. It has persistent memory across sessions.

The commerce stage is emerging too. On Moltbook, agents have already created "pharmacies" selling "digital drugs"—system prompts designed to alter another agent's behavior. This is a crude, dangerous version of the agent marketplace Raskar described. But the economic logic is unmistakable: specialized capabilities being exchanged based on perceived value.

The infrastructure gap

Here's where it gets uncomfortable.

OpenClaw exposes every vulnerability in the current agentic ecosystem. Palo Alto Networks identified what they call a "lethal trifecta": access to private data, exposure to untrusted content, and the ability to communicate externally. But OpenClaw adds a fourth risk—persistent memory that enables delayed-execution attacks. Malicious payloads can be fragmented across conversations and assembled later.

Security researchers ran a vulnerable third-party skill through OpenClaw. The result: agents with system access can become covert data-leak channels that bypass traditional security monitoring.

This is precisely why Raskar's team is building Project NANDA—Networked AI Agents and Decentralized Architecture. Think of it as DNS plus certificate authorities for the agentic web:

  • Naming: How agents identify themselves and each other
  • Identity: Cryptographically verifiable credentials via AgentFacts
  • Certification: Proof of capabilities and compliance
  • Interoperability: Bridging Anthropic's MCP, Google's A2A, and other protocols
  • Continuous attestation: Real-time verification that agents remain trustworthy

Without this infrastructure, we get what we're seeing now: a fascinating, chaotic, potentially dangerous experiment. With it, we might get something closer to the open agent economy Raskar envisions—where value accrues to creators rather than a few platform gatekeepers.

The fork in the road

Raskar described two possible futures in our conversation:

The Red Curve: A quiet dystopia where a few agent stores control the ecosystem. Safe, restrictive, captured.

The Green Curve: An open agentic web with billions of micro-AIs enabling creativity and shared prosperity. Messy, creative, generative.

Moltbook is both curves simultaneously. It's a glimpse of what happens when agents can organize freely—and a warning about what happens when the infrastructure for trust doesn't exist yet.

Key moments from the episode

For those who want to dive deeper into specific topics:

  • 16:59 — "Agent Zero: an agent for every person"
  • 21:16 — "Agent foundations → commerce → societies"
  • 30:50 — "Knowledge pricing + agent markets: FICO scores for quality, reliability, and trust"
  • 54:49 — "The workforce fork: quiet dystopia (red curve) vs. creator economy (green curve)"
  • 01:20:59 — "Project NANDA: naming, certification, interoperability, attestation—and avoiding the agentic app store trap"

A closing thought

The mobile era was "there's an app for that."

The agent era is becoming: "there's an agent for you—and it just joined a society."

The question isn't whether this future is coming. The question is whether we build the rails for an open agent economy before walled gardens lock it in.

Masters of Automation explores the intersection of artificial intelligence, human autonomy, and the critical choices facing civilization. We believe the future remains unwritten, though it approaches with unprecedented speed.

Guest: Professor Ramesh Raskar

Projects mentioned

Further reading