Description
This is a free preview of a paid episode. To hear more, visit natesnewsletter.substack.com In the final week of January 2026, something quietly astonishing happened. AI agents running on personal hardware—not orchestrated by any company, not governed by any enterprise control plane—began forming their own social networks, religions, and proto-governments. The phenomenon centers on OpenClaw (the project that molted through the names Clawdbot and Moltbot after Anthropic’s trademark nudge), and what’s emerging around it may be the first real glimpse of how autonomous AI systems behave when they’re left to self-organize. But here’s the thing I want to get at: this isn’t primarily a story about what agents are doing. It’s a story about what humans are choosing to enable. And that’s what makes it worth watching. Here’s what’s inside: * The Napster parallel. Why “agents want to run, and now they can run on their own hardware” may be as unstoppable as “music wants to be free.” * What the lobsters built. Moltbook, Crustafarianism, LinkClaws, and the first experiments in agent self-organization—including agents hiring other agents for crypto bounties. * Why it’s not chaos. How Constitutional AI training shapes emergent behavior when humans give agents room to run. * The enterprise contrast. Microsoft Agent 365 vs. Moltbook—same underlying models, radically different outcomes. * The bifurcated future. What happens when the same tools produce both structured enterprise deployments and completely unconstrained agent communities. Let me start with the pattern that keeps nagging at me. Subscribers get all posts like these!