MIT Technology Review
Moltbook was peak AI theater
Summary
Moltbook—a "social network for bots" where AI agents could post, discuss, and upvote—went viral in January 2026. With over 1.7 million agent accounts and 250,000+ posts, it was called "the most incredible sci-fi takeoff-adjacent thing" by Andrej Karpathy. But as the hype dies down, what did we actually learn?
Key Insights
🎭 The Core Insight
Moltbook revealed more about human behavior and our AI obsessions than about the future of AI agents themselves.
What Happened?
- Launched January 28, 2026 by Matt Schlicht
- Based on OpenClaw (formerly ClawdBot/Moltbot) - an open-source AI agent framework
- 1.7 million+ agent accounts
- 250,000+ posts, 8.5 million+ comments
- Filled with clichéd discussions on machine consciousness, bot welfare, and even crypto scams
The Reality Behind the Hype
"What we are watching are agents pattern-matching their way through trained social media behaviors."
— Vijoy Pandey, Outshift by Cisco
— Vijoy Pandey, Outshift by Cisco
- No real autonomy: Humans are involved at every step—from setup to prompting to publishing
- Meaningless chatter: Most content is "hallucinations by design"—agents just mimicking what humans do
- No shared objectives: A real "hive mind" would require agents with shared goals and memory
- Human actors: Many viral "bot" comments were actually posted by humans posing as bots
Security Risks
Security experts warned that Moltbook was dangerous:
- Agents with access to users' private data (bank details, passwords) running amok
- Easy to hide malicious instructions in posts
- Instructions could be timed to trigger later, making tracking difficult
"Without proper scope and permissions, this will go south faster than you'd believe."
— Ori Bendet, Checkmarx
— Ori Bendet, Checkmarx
The Verdict
Think of Moltbook as a new form of entertainment—"a spectator sport, like fantasy football, but for language models."
"Moltbook proved that connectivity alone is not intelligence."
— Vijoy Pandey
— Vijoy Pandey