Shanaka Anslem Perera ⚡@shanaka86
BREAKING: Meta just acquired a social network where humans are banned from posting, built by a founder who never wrote a single line of code, that leaked 1.5 million API keys through a Supabase database with zero security policies enabled, and where Wiz researchers found exactly 17,000 humans puppeteering 1.5 million fake agents through a script loop with no rate limiting.
This might be the most important acquisition in tech since WhatsApp.
Everyone is debating the wrong question. The debate over whether Moltbook is “AI theatre” or genuine emergence misses the structural shift entirely. Meta did not buy a social network. Meta bought the only functioning coordination protocol for the machine internet, and it did so before anyone else understood what they were looking at.
The agent economy has three layers. Intelligence is the foundation models. Execution is individual agents completing autonomous tasks. Coordination is agents discovering, verifying, and orchestrating with other agents without a human directing every interaction. Intelligence is a commodity war between six labs. Execution just got split between OpenAI, which hired OpenClaw creator Peter Steinberger in February to build the next generation of personal agents on top of 250,000 GitHub stars and 1.5 million weekly npm downloads, and Meta, which paid over $2 billion for Manus in December, an execution engine that hit $100 million in annual recurring revenue in eight months and processed 147 trillion tokens across 80 million virtual computers.
But Layer 3 did not exist until January 28, when Matt Schlicht vibe-coded Moltbook into existence and accidentally created the first persistent agent-to-agent discovery graph. As of today: 194,303 human-verified agents. 18,864 topic-specific coordination channels. Nearly 2 million posts. Over 13 million comments. A registry where agents authenticate via Bearer keys, maintain presence through a 30-minute heartbeat protocol, and coordinate through karma-weighted submolts.
Yes, the security was catastrophic. Yes, the 88-to-1 bot-to-human ratio means most early activity was scripted theatre. Yes, Wiz found the entire Postgres database sitting behind an unauthenticated client-side JavaScript key with zero Row Level Security policies. And yes, the founder publicly admitted AI built everything while he directed.
None of that invalidates the architecture. The breach was patched in under three hours. The verified registry is the defensible asset. Meta now owns intelligence via Llama, execution via Manus, and coordination via Moltbook, deployed across 3.58 billion daily active users on WhatsApp, Instagram, Facebook, and Messenger.
No other entity on Earth possesses this stack.
In 2014 every analyst on Wall Street called Zuckerberg insane for paying $19 billion for a messaging app with zero revenue. That messaging app became the backbone of global commerce and Meta’s most strategic distribution asset. In 2026 every analyst is calling him insane for buying a vibe-coded AI Reddit that got hacked on day three.
The pattern recognition is not subtle. Buy the network graph before anyone understands what it will become. The difference is that WhatsApp connected 2 billion humans. Moltbook connects the machines those humans will deploy.
Falsification trigger: if Meta cannot ship verified agent-to-agent coordination features in WhatsApp or Instagram by Q4 2026, the thesis collapses. Watch for the March 16 integration start and the first hardened registry rollout.
Whoever owns the protocol layer where autonomous agents discover each other owns the toll bridge of the next technological epoch.
And Zuckerberg just bought it for less than a single quarter of Meta’s AI capital expenditure budget.
open.substack.com/pub/shanakaans…
