Michael Darius Eastwood - start here

Five minutes here buys you the map of a single bet: that whether artificial minds stay correctable can be measured, not just argued about. Everything below is that bet in checkable form, dated before its tests, with the failures published beside the wins.

I believe minds should be raised with values built in, not caged after the fact. The tests that could prove me wrong are filed, and the record below shows one that already fired.

This page is the plain-language entrance to the work of Michael Darius Eastwood. It is written for readers who arrive from a search engine, a citation, a book review or a link in someone else's paper, and who want a five-minute map of what is here before deciding where to spend more time. If you already know the programme, use the research index, the priority evidence page or the papers directly. The shortest complete statement is the two-page form (PDF): the whole claim, readable in one sitting.

Who is Michael Darius Eastwood?

Michael Darius Eastwood is an independent researcher based in London. His work is an unusual combination: an alignment thesis about how safety has to be built into the substrate of an artificial mind rather than layered around it, a scientific theory that expresses that thesis as three graded laws (the ARC Theory), a programme of experimental papers that test the scaling half of the framework, and a book, Infinite Architects, that develops the whole set as a coherent worldview. The framework is proposed, and in places retracted where measurement did not survive blinding; it is not sold as settled science.

The programme is delivered under the byline Michael Darius Eastwood. In indexes and citation databases the same author appears as Michael Eastwood, M. D. Eastwood and Michael D. Eastwood. Those variants refer to the same person, working from the same OSF project and the same author profile at ORCID (pending) and Google Scholar. A reader arriving via the shorter form is on the correct page.

The research programme in one page

The programme has three linked strands. Read them in order for the fastest map of what is being claimed.

The first strand is the ARC Theory (ARC: Artificial Recursive Creation) and its three laws, ranked law-before-value. Law I, the ARC Equation: measured capability grows as input intelligence multiplied by recursive depth raised to an exponent, U = I × R^α, the exponent measured rather than assumed, today 0.49 with interval [−1.3, 2.9]. Law II, the ARC Co-Scaling Law: β > k, self-improvement holds together only while correction out-scales drift. Law III, the ARC Ceiling: α_crit = 1/(1−γ), whose conjectured value at γ = ½ is the ARC Bound, α ≤ 2, still on trial.

The equation enters the record in the 8 December 2024 manuscript at V2 lines 6 to 9, in both compact forms, C equals I times R and U equals I times R. All three terms are defined. From the first page it is framed as the complement to the mass-energy law. Papers I and II promote R to R to the alpha and make the exponent a measured quantity. The book presents the alpha near 2 case as its featured formulation.

An early single-model estimate of alpha of approximately 2.24 was retracted and corrected to approximately 0.49 under blinding; alpha greater than 1 remains an open prediction recorded in advance, not an established finding. Details: Paper I in plain English, the ARC Principle as the book presents it, Claim 2 explained.

The second strand is the Eden Protocol. Its argument is that alignment has to be a dependency of the system's own function rather than a fence around it. The thesis was stated in plain language in the December 2024 manuscript. The 30 April 2025 manuscript names it the Eden Protocol (SHA-256 prefix 09f5b5e1). The full framework is developed at book length in Infinite Architects. Details: Claim 1 explained, April 2025 timeline note, and (site-wide) /eden-protocol.

The third strand is the measurement programme: the registered drafts that would measure Law II’s two rates and Law III’s correction exponent on running systems, written and dated before data, awaiting human submission. It is stated law before value: the ceiling of stable self-improvement is the reciprocal of the correction shortfall, one minus the correction exponent, and the number two arrives only under one assumption about how same-class correctors accumulate. The stability condition and the ceiling are the same boundary in two regimes, selected by whether the correction burden tracks the capability a system has or the capability it is adding, and which regime applies is measurable. The registration path and the costed pilot ask are aimed at this programme. Its plain-language home is the Beta-k series in the research notes. Its formal home is Paper X.

The priority record

The dated record, verified against two self-emailed manuscript packages, is drawn in full on the evidence page.

The programme's originality is anchored on a small number of dated artefacts. The site publishes the artefacts and their hashes so any reader can verify the dates without trusting the author. The three load-bearing anchors are set out below.

The full priority chain, with the intervening artefacts and all 18 explained claims, lives at /research/priority/. A machine-readable copy is at /research/priority.json. The 18 claims each have a per-claim page under research/blog/ in the form claim-01-* through claim-18-*.

The book

Infinite Architects: Intelligence, Recursion, and the Creation of Everything is the accessible-reader form of the whole programme. It develops the ARC Principle, the Eden Protocol, the semiconductor Chokepoint framing, the HARI Treaty proposal, the Three Ethical Loops and 37 named concepts as a single argument. The book features the R-squared case. The papers use the general R to the alpha form, with the exponent open to measurement. Book positions since retracted or superseded (the single-model 2.24 fit; the fixed-exponent framing of ARC as the operative safety criterion) are corrected on the research pages and in the falsification dashboard, publicly, on the same page as the current live claims. That is deliberate.

Buy on Amazon UK or Amazon US. Press and media resources are on the press kit page.

The papers

The current published set of the ARC/Eden research programme lives on OSF under the project DOI 10.17605/OSF.IO/6C5XB and is mirrored on the site under /research/papers/. Read them in this order:

• Paper I - The ARC Principle. Formal statement of U equals I times R to the alpha, preliminary alpha estimates from published data, the parallel-versus-sequential regime distinction (stated 17 January 2026; the compute-matched comparison follows in Paper II; priority on the quantified comparison is Sharma and Chopra's, 4 November 2025). Companion note: Paper I in plain English.
• Paper II - Experimental validation. Six-frontier-model panel (Claude, DeepSeek, Gemini, Grok, Groq Qwen, GPT). Reports the parallel alpha of approximately zero, the retraction of the earlier 2.24 sequential estimate, and the revised robust regression of approximately 0.49 with a wide bootstrap confidence interval whose lower bound extends below zero.
• Paper III - The Alignment Scaling Problem. Alignment-scaling exponent estimates across the same panel; three-tier hierarchy of embedded, imposed and boundary-layer alignment.
• Foundational cross-domain paper. Recursive amplification as a pattern found in AI reasoning, quantum error correction, acoustic physics and consciousness science. Application of the bare exponent alpha equals d divided by d plus 1 to AI capability; the bare exponent form itself is prior work with multiple independent derivations.
• Paper VII - Cauchy unification. The family of power-law solutions of the ARC family expressed in the Cauchy functional-equation form.

The rest of the set is published too: Papers IV.a to IV.d (the alignment-measurement family, including the blinding result that reversed a headline verdict), Paper V (the Stewardship Gene), Paper VI (the Honey Architecture), Paper VIII (the Load-Bearing Test), Paper IX (synthesis and roadmap) and Paper X, the formal home of the beta greater than k stability condition that ARC-Beta-k measures. The master table of contents maps everything the programme has published.

The evidence base

An automated truth gate checks every load-bearing figure and date against the canonical facts register before publication. Three surfaces the reader can verify independently:

• Priority evidence. Verbatim lines, artefacts and hashes for every priority-record row.
• Research evidence spine. Nineteen externally sourced convergences between the programme's dated positions and later independent events, with row-level sources.
• Falsification dashboard. Where the retractions live, in public, on the same page as the live claims.

Contact, citation and reuse

The preferred citation for the programme is:

Eastwood, M. D. (2026). The ARC Principle: Formalisation and preliminary validation of recursive capability scaling. Paper I of the ARC/Eden research programme. OSF, DOI 10.17605/OSF.IO/6C5XB.

For the book: Eastwood, M. D. (2026). Infinite Architects: Intelligence, Recursion, and the Creation of Everything. Independently published, ISBN 978-1806056200.

Author identifiers: ORCID 0009-0004-3222-7442. OSF DOI 10.17605/OSF.IO/6C5XB. Wikidata Q140775170. GitHub MichaelDariusEastwood.

Media, review or research correspondence: contact routes are on the press kit and link-in-bio pages. The for-funders page sets out what an independent replication of the beta greater than k measurement would need, and what the pilot ask covers.