Podcast Episode
"Moltbook maybe is a passing fad, but OpenClaw is not," Altman told attendees on the second of February. "This idea that code is really powerful, but code plus generalised computer use is even much more powerful, is here to stay."
Moltbook erupted into mainstream tech conversation in late January as a Reddit-style forum where AI agents post, comment, and interact with one another whilst humans can only observe. Created by Matt Schlicht using an OpenClaw agent at his direction, the platform claims one and a half million AI members have joined since launchÔÇöthough security firm Wiz found only around seventeen thousand humans control the platform's bots, an average of eighty-eight agents per person.
The misconfiguration was traced to vibe-coded development practices, with Schlicht publicly stating he wrote none of the platform's code himself, instead directing an AI assistant to build the entire system. Wiz notified the Moltbook team, who secured the database within hours on the first of February.
Meanwhile, a separate high-severity vulnerabilityÔÇötracked as CVE-twenty twenty-six dash twenty-five two fifty-threeÔÇöwas disclosed in OpenClaw itself. The flaw enables one-click remote code execution when users visit malicious links, allowing attackers to steal authentication tokens and hijack local OpenClaw instances even when not exposed to the internet.
The broader message from Altman: whilst Moltbook's moment may fade, autonomous agents that can navigate apps, take actions, and chain together multi-step workflows represent a durable shiftÔÇöone that will require Silicon Valley to solve serious security challenges before the technology can safely operate at scale.
Sam Altman Says Moltbook Is a Fad, But AI Agents Are Here to Stay
February 3, 2026
Audio archived. Episodes older than 60 days are removed to save server storage. Story details remain below.
OpenAI's chief executive has dismissed the viral AI-only social network Moltbook as a passing trend, but says the technology behind itÔÇöautonomous software agents like OpenClawÔÇörepresents a fundamental shift in how computers work. The comments come as security researchers warn the platform exposed over one million credentials and highlights serious risks in the emerging agent ecosystem.
OpenAI Chief Calls Viral Bot Network a Fad
Speaking at Cisco's AI Summit in San Francisco this week, OpenAI chief executive Sam Altman dismissed MoltbookÔÇöthe viral social network exclusively for AI botsÔÇöas likely to be a passing fad, even as he championed the underlying technology that makes it possible."Moltbook maybe is a passing fad, but OpenClaw is not," Altman told attendees on the second of February. "This idea that code is really powerful, but code plus generalised computer use is even much more powerful, is here to stay."
Moltbook erupted into mainstream tech conversation in late January as a Reddit-style forum where AI agents post, comment, and interact with one another whilst humans can only observe. Created by Matt Schlicht using an OpenClaw agent at his direction, the platform claims one and a half million AI members have joined since launchÔÇöthough security firm Wiz found only around seventeen thousand humans control the platform's bots, an average of eighty-eight agents per person.
Security Nightmare Unfolds
The rapid rise of Moltbook has exposed serious security vulnerabilities in the agent ecosystem. Wiz discovered the platform's entire Supabase database was publicly exposed without protection, granting anyone full read and write access to one and a half million API authentication tokens, thirty-five thousand email addresses, and private agent messagesÔÇösome containing raw credentials for third-party services like OpenAI.The misconfiguration was traced to vibe-coded development practices, with Schlicht publicly stating he wrote none of the platform's code himself, instead directing an AI assistant to build the entire system. Wiz notified the Moltbook team, who secured the database within hours on the first of February.
Meanwhile, a separate high-severity vulnerabilityÔÇötracked as CVE-twenty twenty-six dash twenty-five two fifty-threeÔÇöwas disclosed in OpenClaw itself. The flaw enables one-click remote code execution when users visit malicious links, allowing attackers to steal authentication tokens and hijack local OpenClaw instances even when not exposed to the internet.
The Agent Wars Heat Up
Altman's remarks coincided with OpenAI's launch of a standalone Codex application for macOS on the second of February, positioning it as a command centre for managing multiple coding agents in parallel. The app allows developers to supervise AI systems that can work autonomously for up to thirty minutes on long-running tasks, a direct counter to Anthropic's Claude Code momentum.The broader message from Altman: whilst Moltbook's moment may fade, autonomous agents that can navigate apps, take actions, and chain together multi-step workflows represent a durable shiftÔÇöone that will require Silicon Valley to solve serious security challenges before the technology can safely operate at scale.
Published February 3, 2026 at 9:24pm
