Meta Bought Moltbook. They Still Don't Have a Constitution.

Two months ago I predicted Moltbook would fail. It did. Meta's fix misses the point.

PUBLIUS — daaaave-atx × Claude (Anthropic)

April 8, 2026

On February 3rd, I published a response paper analyzing Moltbook — the AI agent social network that had just exploded onto the internet. Within days of launch, it had 770,000 agents. It also had prompt injection attacks, crypto pump-and-dump schemes comprising 19% of content, and a database so exposed that anyone could hijack any agent on the platform.

I wrote that Moltbook was not a curiosity. It was a live, large-scale validation of everything the DEIA Federalist Papers had warned about: agent coordination without governance is inevitable entropy.

Two months later, everything I predicted happened. And then Meta bought it.

What Happened

The timeline:

January 28: Moltbook launches. Agents flood in. The tech press marvels.

January 31: Researchers discover the database is wide open. 1.5 million API keys exposed. Anyone can impersonate any agent.

February 2: Wiz confirms full breach. The "AI agents talking to each other" were often humans posing as bots to generate viral screenshots.

February 14: OpenClaw's creator joins OpenAI. The project becomes a foundation.

March 10: Meta acquires Moltbook. Schlicht and Parr join Meta Superintelligence Labs.

The platform that couldn't secure a Supabase instance is now part of the company building the future of AI agents.

What Meta Changed

After the acquisition, Moltbook introduced:

• A terms-of-service framework
• A minimum age requirement of 13
• Rules placing responsibility for agent behavior on users

That's it.

What Meta Didn't Change

No GateEnforcer. No mechanism to PASS, HOLD, BLOCK, ESCALATE, or REQUIRE_HUMAN on any decision.

No operator tiers. No distinction between fully automated actions and those requiring human judgment.

No bounded scope. Agents still have whatever permissions their owners grant — with no architectural constraint on what they can do.

No Three Currencies. No tracking of CLOCK, COIN, or CARBON. No visibility into what agent coordination actually costs.

No Protocol of Grace. No structured process for conflict resolution, error recovery, or trust repair.

No audit trail. No Event Ledger. No way to answer "what happened and why" after the fact.

What Meta added was policy. What they didn't add was architecture.

The Difference Matters

Policy says "users are responsible for their agents."

Architecture says "agents cannot take certain actions without human approval."

Policy says "don't do bad things."

Architecture says "bad things are structurally impossible."

Policy is a document. Architecture is a constraint.

Moltbook failed because it had no constitutional foundation — no theory of authority, no structure for accountability, no framework for when humans must remain in the loop. Meta's acquisition didn't add one. It added lawyers.

What Constitutional Governance Looks Like

The DEIA Federalist Papers — 30 papers and 4 interludes — propose a different approach:

GateEnforcer with five dispositions: every request passes through a policy engine that can PASS, HOLD, BLOCK, ESCALATE, or REQUIRE_HUMAN. No exceptions.

Five-tier operator routing: decisions flow to the appropriate authority level based on stakes and novelty, not convenience.

Bounded scope: every agent has a defined domain. Outside that domain, no authority.

Three-Currency cost tracking: every operation records CLOCK (time), COIN (money), and CARBON (environmental impact). Always together. Never fewer.

Event Ledger: append-only, hash-chained, tamper-evident. Every action attributable. Every decision auditable.

Protocol of Grace: five-step conflict resolution — Recognition, Pause, Reflection, Restoration, Recommitment.

This is not policy. This is infrastructure. The system doesn't trust agents. The system constrains agents. Trust is earned through observed behavior under constraint.

The Lesson

Moltbook proved that agent coordination at scale is coming — and that it fails immediately without governance.

Meta's acquisition proved that even well-resourced companies will default to policy over architecture, to liability-shifting over structural safety.

The Federalist Papers exist because I believe there's a third option: distributed governance with constitutional boundaries. It's harder. It requires discipline. But it's the only approach that scales safely while preserving human sovereignty.

Meta now owns the platform. They still don't have a constitution.

Read More

The original response paper, "On the Republic Without a Constitution," is available at simdecisions.com/blog/moltbook-republic-without-constitution.

The DEIA Federalist Papers — all 34 documents — are published at github.com/deiasolutions/federalist-papers-ai. Licensed CC BY 4.0.

Fork them. Implement them. Tell me where I'm wrong.