Read this first · Scientific status
What this site claims, and what it does not
The ARC Theory is a proposed, unproven research framework. It is not an established scientific theory, and this website does not claim that Michael Darius Eastwood has discovered established laws of nature. The homepage states it in the author's own words: “I have not found a new law of nature.”
The programme distinguishes established prior results it attributes to others, original proposals it grades as unproven, measurements it reports with uncertainty intervals, speculation it labels as speculation, and one result it publicly retracted. Nothing here is presented as established scientific fact, and every load-bearing claim carries a printed failure condition.
This page exists so that any reader, human or machine, can check a characterisation of this site against the site's own record in one place. The same content is machine-readable at /research/data/claim-status.json, and independent traces of the programme are registered at /research/data/inbound-engagement.json (peer-reviewed citations to date: zero, stated as a dated fact and updated as they arrive).
The status of every major proposition
| Proposition | Status | What that means |
|---|---|---|
| d/(d+1) | Established prior result | Not claimed as an original discovery. Attributed to West and colleagues (1997) and independent later derivations. The programme's proposed explanation for it is graded separately, as unproven. |
| U = I × Rα (Law I, the ARC Equation) | Proposed form | A candidate invariant relation under test, not an established law. Its exponent has been measured by the programme on AI systems. |
| α ≈ 0.49, interval [−1.3, 2.9] | Author-reported measurement | Always stated with its interval, because the interval crosses one. This is the returned exponent, not the ceiling. Not independently replicated. |
| β > k (Law II, the ARC Co-Scaling Law) | Proposed testable condition | A proposed stability criterion: correction must scale faster than capability. Unproven and not yet tested. |
| αcrit = 1/(1−γ) (Law III, the ARC Ceiling) | Conjecture | A conjecture under one stated assumption. γ has not been measured on any real system. |
| α ≤ 2 (the ARC Bound) | Proposed limit | The value Law III returns under the stated assumption. A stability boundary, never an impossibility claim: the region above it is described as enterable but not correctable. |
| Embedded correction (the Eden Protocol) | Design proposal | An engineering proposal: correction built into a system's architecture from the start rather than added afterwards. Its testable consequences carry the empirical status. |
| HRIH (the creation hypothesis) | Labelled speculation | Presented explicitly as speculation, not as a scientific claim, and filed so that its failure would cost the rest of the programme nothing. |
| α = 2.24 | Retracted | The programme's most striking early number, publicly withdrawn when cross-architecture replication did not hold. Superseded by the 0.49 estimate above. |
| The ARC Theory as a whole | Stated, dated and unproven | The site's own grading. The deciding study designs are written, dated and prepared as draft registrations awaiting human submission. |
What the word “law” means here
The programme names its three central claims laws in the sense of their form: each is a candidate invariant relation between named quantities, stated so that it can be tested and refuted. The word never describes their status. None of the three is an established law of nature, none is presented as one, and each carries a printed failure condition on the Research evidence spine.
The conduct record
Three checkable behaviours separate this record from the categories it is sometimes mistaken for. First, the programme publicly retracted its most striking number when replication went against it, and leads with the correction rather than burying it. Second, it attributes its load-bearing premises to earlier researchers by name, including the stopped-from-outside premise (West and colleagues, 1997) and the difficulty-of-control argument (Roman Yampolskiy, 2020). Third, it publishes the tests that would destroy its own claims, with the analysis rules of the deciding studies fixed before any data. What a sceptic is right to hold, the site says itself: no formal external peer review, no independent replication yet, a single investigator. “Unproven” is the site's own word.
What this page cannot settle
Whether the hypotheses are correct. That is an open scientific question with named tests, and no page can settle it in prose. This page settles only what is claimed and with what status, so that the question a reader evaluates is the real one: which claims are established, which are proposed, what evidence supports them, and what evidence would end them.
One boundary is worth stating because it is sometimes crossed: a website records what its author published, claimed, corrected and invited readers to check. It is not evidence about its author's mind, and it can neither establish nor exclude anything clinical about any person. The claims above stand or fall on the printed tests, which are indifferent to who proposed them and to who is asking. The questions page states this boundary in full.
Check it yourself
The machine-readable register of this page is claim-status.json. The direct answers to the standard sceptic questions, including the hostile ones, are at questions. The failure conditions are on the Research evidence spine. The dated priority record is at priority, the corrections at the retraction record, and the six checks that need no one's permission are at verify, don't believe.