The Disagreement — Summary

Two identical households asking two advisors the same question in 2015 got different answers, and the profession’s embarrassment, disagreement, had one unpriced property: it decorrelated the behavior of millions of similar households. Refinancings smeared across quarters, reallocations offset, exits staggered. The defect was a stabilizer, and nobody built it.

Models converge. Trained on overlapping corpora toward the same published consensus, the optimizing instruments give similar households similar answers at similar moments, and the answers execute by default. The claim: convergent advice at household scale is the coherent-crash mechanism running through ten million ledgers instead of ten thousand funds, the same refinance week, the same rotation, arriving as one correlated event no individual ledger did anything wrong to produce.

The premise’s status is declared in the body: asserted, not measured. No study compares the dispersion of human advisory recommendations against model recommendations for matched households, and the missing study is named precisely, then states why surfacing precedes measurement: requiring proof before a risk can be named suppresses exactly the risks nobody has studied. The claim is dated to end-2030 with its miss condition stated: if the study shows model dispersion comparable to the human baseline, the premise fails and the claim stands as a scenario that did not arrive.

Conditional on the premise: dispersion’s removal converts noise into systematic exposure with no responsible party; synchronization is legible and therefore harvestable, a new arbitrage created by closing the old ones; and deliberate de-correlation is trivial, unrewarded, and owned by no one. Decision rule: hold nothing whose solvency assumes household behavior remains staggered, until the study exists and says otherwise.