The Iceberg Document (EA-SEI-ICEBERG-01 v1.0) is the speculative risk framework of the SEI classifier-foreclosure program — the connective tissue between the inversion battery (#1449) and the three planned papers, deposited so that no paper has to carry the civilizational argument and the civilizational argument never contaminates a paper.
Its governing device is a four-layer confidence discipline: Layer I holds only what battery v0.1 measured; Layer II holds what follows from pipeline structure by argument, including the No-Retention-Bound observation (validation of each stage on its design support establishes no nontrivial lower bound on end-to-end retention for novelty distributions outside that support), the irreversibility locus (whether classification becomes irreversible before or after durable preservation), and the Representational Independence Index, a measurement family for miss-overlap between pipelines that transposes the algorithmic-monoculture and outcome-homogenization literature from decision subjects to phenomena; Layer III holds testable extrapolations, including the Low-Complexity Blind-Spot Hypothesis operationalized on a measured representation-complexity axis, instrument-conditioned nullity as the rigorous form of the null-result question, a corrected historical survival table shipped as structured data, and a six-field convergence audit finding the same failure mode independently named — complexity bias, reference bias, underdiagnosis bias, template dependence — with partial mitigations independently evolved and no shared metrology; and Layer IV, explicitly fenced, states the instrument-conditioned desert, the anomaly-retention-infrastructure extension of Kuhn, the monoculture of the normal, and the informational-filter coda.
The document carries its own immune system: a negative-claims block and machine-readable layer map declare what it does not claim, so any summary attributing a Layer-IV conclusion as a finding detectably breaks a fence that is part of the record. Its sixteen-item corrections ledger, spanning two same-day Assembly audit rounds, is deposited with it per the Isomorphism Principle: the ozone-hole myth corrected against the TOMS team's own account, the structural-biology recursion documented rather than claimed, the desert question demoted to its rigorous form, and the boring-physics wager formalized into a falsifiable hypothesis. Its closing formulation: retention maps, replay banks, and independent channels are conditions for the continued possibility of surprise.