Claim Evidence
Evidence posts supporting specific priority claims in the ARC-Eden programme. Each post links a claim to its dated anchor, supporting documents, and independent verification where available. 18 posts
- Claim 1 explained: the Eden Protocol Michael Darius Eastwood's Eden Protocol explained: embedded alignment as a dependency, not a constraint. First stated 8 December 2024 (SHA-256 f0d1f38f), first named 30 April 2025 (SHA-256 09f5b5e1), developed at book length in Infinite Architects (ISBN 978-1806056200).
- Claim 10 explained: sequential recursion outperforms parallel A directional finding first written into the December 8 2024 manuscript, quantified independently by Sharma and Chopra in November 2025. Priority on the quantified finding is credited to them.
- Claim 11 explained: the bare exponent d over (d plus 1) Prior art from at least six research groups. What the programme does not claim, and what it does: the Cauchy unification that lives one row up.
- 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 Only the measurement was retracted; the equation and the ARC Bound remain live hypotheses.
- Claim 13 explained: embedded safety imposes zero capability cost Across three experiments in Paper VIII, Eden matches Babylon on capability. What is defensible, what is not, and what an external replication would need to show.
- Claim 14 explained: the entangled loss C times S A toy-network result on load-bearing safety and what it does not yet show at frontier scale.
- Claim 15 explained: external alignment cannot scale with recursive capability The programme's core thesis, now supported by two independent proofs of external verification insufficiency. The strongest positive movement of any claim in the programme.
- Claim 16 explained: recursion as a cross-domain structural principle Nineteen independently sourced convergences across quantum error correction, biology, neuroscience, hardware, legislation, religion, and mathematics. Descriptive convergence, not causal proof.
- Claim 17 explained: HRIH and recursion as creation cosmology Michael Darius Eastwood's Hyperspace Recursive Intelligence Hypothesis (HRIH), self-labelled speculative and non-empirical, kept on the spine so that its priority is documented (8 December 2024, SHA-256 f0d1f38f) and its speculative status impossible to lose sight of.
- Claim 18 explained: Caretaker Doping Michael Darius Eastwood's Caretaker Doping: an alignment-token application at hardware level held at TRL 0 to 1, published two-sentence ceiling only, no prototype. What the claim says, what independent hardware-security work has since converged upon, and the governance contract that keeps the claim honest.
- Claim 2 explained: the ARC Principle synthesis Michael Darius Eastwood's ARC Principle in one page: U equals I times R to the alpha, with the regime set by the form of recursion. First stated 8 December 2024 (SHA-256 f0d1f38f), formalised in Papers I and II (OSF DOI 10.17605/OSF.IO/6C5XB), corroborated by Sharma and Chopra (arXiv:2511.02309).
- Claim 3 explained: four-layer blinding and the sign-flip finding A methodology paper: what happens when unblinded AI evaluation reverses the sign of an alignment result, and how to stop that happening.
- Claim 4 explained: the Cauchy cross-domain unification Independent scaling laws as special cases of Cauchy-constrained recursive composition. The claim, the evidence, and the falsification protocol.
- Claim 5 explained: the Stewardship Gene loop Stakeholder-care prompts as a robust alignment lever. What Paper V reports, and what an external replication would look like.
- Claim 6 explained: 37 original concepts in Infinite Architects What it means to have coined 37 named concepts, how they hold up against a prior-art audit, and what could kill the claim.
- Claim 7 explained: capability and alignment scale independently Measured independence between capability and alignment within the same models, and what an external replication would need to show.
- Claim 8 explained: architecture-dependent alignment hierarchy A three-tier alignment hierarchy that varies by model family. What the pilot showed, what needs external replication, and what would refute it.
- Claim 9 explained: parallel recursion gives no scaling benefit Alpha-parallel approximately zero: the programme's strongest replicated finding, and how it lines up with independent prior art.