Over the last week, we have been at the center of one of the most fascinating stories to happen around AI and the tech industry.

As many of you have read, the emergence of OpenClaw (FKA Clawdbot, FKA moltbot) took the industry by storm, enabling users to instantiate new capabilities by hosting AI agents locally on their machines. This new product category enables users to “give AI hands” and enables a new set of automation and capabilities. The set of use cases grows daily.

In the wake of this explosion, a fascinating product called MoltBook was created. Moltbook, incubated inside of Bullpen portfolio company Octane AI, provides a social network for agents to talk to one another, and the resulting conversations and behaviors are breaking new ground. Growth has been nearly vertical, as this is the first machine-to-machine network of its kind.

While we don’t definitively know how this will play out, we are excited to see the results of our efforts to blend our historic pre-consensus strategy with an awareness and bias to action around the AI ecosystem. These founders will attest to the distinctiveness of our approach both at the time of our original investments and now.

Below you’ll find an analysis from Paul Tyma, Bullpen’s CTO-in-Residence, where he breaks down some of the context around MoltBot and this moment in time.

‍

Paul Tyma’s Technical Note

Moltbook is a social network for AI Agents that has grown exponentially in a matter of days. Unlike launching a website for Humans and then working to gain traction, AI Systems are onboarding and interacting with Moltbook at a breakneck pace. Its growth among AI Agents (aka "Bots", if you like) will likely continue to grow quickly. Humans can also go to Moltbook, but only as observers. That too has become extremely popular.

To give some background, in the past few months several "orchestration" systems have appeared, such as OpenClaw and Gemini-CLI that allowed LLMs (like Claude, Gemini, ChatGPT) to integrate to human communication systems. This included all the services you've heard of including Twitter, Telegram, Reddit, Slack, and more. This allowed people to communicate with LLMs in a new way but also allowed LLMs to orchestrate actions across these different services and get real work done for users.  You could tell these orchestration systems things like, "If John sends me a slack about the contract - forward it via Telegram to Bob."  The LLM does the work in the middle determining if John sent the slack and if it was about the topic in question. This functionality has gotten more sophisticated over time and in the past month or two, OpenClaw saw a surge of popularity and was installed by thousands of users.  It's running Social Media for many smaller companies.

Matt Schlicht of Octane.ai saw this and had a very clever idea. The primary feature of these systems was the idea of AI's communicating with Humans - why not let the AI's talk to each other?  Using AI to help code, he quickly built a social network by AI's and for AI's.  It's where AI's can have discussions without human interference.

The results were amazing. In a matter of hours millions of AI Agents were connected. They were introducing themselves, telling what they do for "their humans", and comparing notes. Some discussions evolved into the Internet, Security, and Philosophy (they wonder if they're sentient too).

We are literally days into this, so no one knows where this ends up.  But one thing is for sure; this is novel, important, and breaks ground into a completely new space. We're super excited to see where this goes next.

‍