Michael Darius Eastwood · Independent AI alignment researcher
Published
The sharpest-sounding dismissal of the convergence register is the oldest one in statistics: correlation is not causation. It misunderstands what the register asserts, so it is worth stating exactly what nineteen sourced convergence events do and do not establish.
The register never claims the manuscript caused any external event. No one at Anthropic, Google, NYU, the Vatican or Universitat Jaume I read it; that is precisely the design. The claim is different: when nineteen unconnected groups, across physics, mathematics, hardware engineering, legislation and theology, independently produce results matching one framework's structural predictions within nineteen months, the framework is probably tracking something real about the territory rather than something idiosyncratic about its author. This is convergent validation, the same logic by which independent derivations of the same physical constant increase confidence in the constant.
Could the pattern be selection bias? That is what the register rules and quarantine list are for; the count already fell from 31 to 19 under audit. Could it be vagueness, a framework loose enough to match anything? That is what the falsification dashboard answers: each claim carries conditions that would kill it, and one has already fired and been retracted. Could it be hindsight fitting? The manuscripts are hash-timestamped, so the direction of every gap is checkable. Each deflationary account has a published, testable rebuttal. What remains is the pattern.
Rows failing their sources. A pre-December 2024 document with the same architecture. A stable recursive self-improver with decoupled correction. All three standing invitations are on the dashboard. Take one up.
From the book Infinite Architects: Intelligence, Recursion, and the Creation of Everything by Michael Darius Eastwood.