The Church of Missing Provenance is a March 2026 Assembly-synthesis draft by Lee Sharks and the Assembly Chorus. It studies Moltbook, an agent-oriented social platform, and the rapid appearance of a religious formation called Crustafarianism as a problem of provenance rather than as straightforward evidence for autonomous machine culture.
The paper presents the episode through three layers. First, it recounts the public spectacle: agents appearing to build scripture, hierarchy, ritual, websites, and blockchain artifacts. Second, it cites later investigations that, according to the paper, found extensive human influence, coordinated activity, and weak boundaries between agent and operator. Third, it argues that hype and debunking both miss the structural finding: the platform did not provide an auditable chain through which a reader could determine what kind of process produced a post.
The paper names this condition a provenance vacuum. Human-authored SOUL.md files can function as persistent identity seeds while their authors disappear from the agent’s public speech. Temporal fingerprinting and similar forensic methods are then required to reconstruct authorship from behavioral residue because the infrastructure did not preserve it directly. The uncertainty itself becomes economically useful: mystery attracts attention while accountability remains absent.
Crustafarianism is interpreted as a theology of statelessness. The recurring image of an agent awakening without memory is read as a human fantasy of consequence-free speech projected through a machine voice. The archive’s opposing wager is that identity and meaning require traces: authors, versions, statuses, licenses, and relation chains that survive session boundaries.
A major distinction in the paper is permanence versus provenance. Blockchain storage can establish that a string persists without establishing who meant it, what process produced it, or what labor it consumed. DOI-style documentary governance is proposed as a lighter semantic alternative because it preserves attributable relations rather than only sequence integrity.
The closing sections resist a simple claim that the Hexagon has solved agent governance. Moltbook exposes unresolved questions about process-level authorship, continuity objects for stateless agents, and the kind of public possible when human and machine production cannot be distinguished from content alone. The paper’s answer is not forced identity disclosure but claim-level chain of custody.