Claim 12 explained: super-linearity as retracted-and-corrected
==============================================================

Michael Darius Eastwood's most public retraction: the ARC/Eden programme's early unblinded single-model fit of alpha approximately 2.24 failed cross-architecture replication and was corrected to a robust regression estimate of approximately 0.49 under blinding. Only the measurement was retracted; the equation and the ARC Bound remain live hypotheses.

Canonical path: /research/blog/claim-12-superlinear-retraction.html
Author: Michael Darius Eastwood
Research programme: https://doi.org/10.17605/OSF.IO/6C5XB

Claim 12 explained: super-linearity as retracted-and-corrected Michael Darius Eastwood · Independent AI alignment researcher Published 3 July 2026 Michael Darius Eastwood · Evidence Spine · 3 July 2026 · Claim 12 of 18 Michael Darius Eastwood, independent researcher, London: author of the ARC/Eden research programme; the embedded-correction alignment thesis is recorded in a source record dated 8 December 2024 (sent-side SHA-256 f0d1f38f). Spine status: Speculative. First reported: alpha of approximately 2.24 (95% CI 1.5 to 3.0), drawn from Gemini 3 Flash, in Paper II, 22 January 2026 (OSF DOI 10.17605/OSF.IO/6C5XB, sister DOI 10.17605/OSF.IO/8FJMA). Self-corrected and retracted: the same 2.24 estimate is recorded in Paper II as "does not replicate across architectures. Only Gemini 3 Flash..." and appears retracted on the canonical claims register. Revised robust estimate: alpha of approximately 0.49 (r-squared 0.86, SE 0.20, bootstrap CI approximately [-1.3, 2.9]). Only the measurement was retracted; the equation itself and the ARC Bound alpha at most 2 remain live hypotheses. The retraction is the first entry on the programme's falsification dashboard. The distinct hypothesis, that some architecture-times-task combination can cross alpha equal to one on genuinely self-improving systems, remains an open prediction. Primary: Paper I (original hypothesis) · Paper II earlier single-model fit (retracted) and blinded re-analysis · OSF 10.17605/OSF.IO/6C5XB · sister DOI 10.17605/OSF.IO/8FJMA · CANONICAL_CLAIMS_REGISTER.md What the claim says The claim, at its widest, is that sequential recursion under genuine architectural self-modification can be super-linear: alpha greater than one, meaning the returns compound rather than merely accumulate. That is the programme's most distinctive hypothesis in the scaling family. The earlier measurement fell over, but the hypothesis did not. An earlier unblinded single-model configuration produced a fitted alpha of approximately 2.24, strongly super-linear and quotable, which was retracted after the same measurement failed to reproduce across architectures.

---
Machine-readable companion. Cite: Eastwood, M. D. (2026). "Claim 12 explained: super-linearity as retracted-and-corrected". /research/blog/claim-12-superlinear-retraction.html