Objections and Rebuttals
Honest engagement with objections to the ARC Principle, the Eden Protocol, and the wider research programme. Every objection is stated in its strongest form before the evidence-based rebuttal. 19 posts
- 'A self-email proves nothing.' What a Gmail server timestamp actually establishes. What Google's infrastructure does and does not prove about a December 2024 manuscript, stated precisely.
- 'AI wrote this.' The evidence chain does not depend on the drafting method. Why models cannot forge hashes, retract old numbers, or run kill-conditions, and what that leaves the author responsible for.
- 'Books are not science.' Correct. Here is what the book is doing on the evidence page. The distinction between a proof and a timestamped artefact, and why the January 2026 book is only the latter.
- 'Correlation is not causation.' The register does not claim causation. That is the point. What nineteen independent convergences do and do not establish, in plain terms.
- 'Everyone knew AI was risky in 2024.' Fine. Show me the document. The most common objection to the December 2024 priority claim, and the thirty-second test that settles it.
- 'Hardware ethics is impossible.' The industry disagrees, on the record. Three published references, one issued patent set, and what they collectively concede about embedded-substrate constraint on AI compute.
- 'Independent researcher means crank.' The base rate is real. Here is how the record distinguishes itself. The uncomfortable statistics of independent AI-risk writing, and the specific machinery that separates a record from a rant.
- 'Sharma and Chopra were first.' Yes. The register says so in bold. Why the programme credits priority on sequential versus parallel scaling to other researchers, and what remains its own.
- 'The 78 per cent figure is misleading.' The RL-training condition, explained properly. What Anthropic's alignment faking paper actually measured, and why this programme qualifies the number every time.
- 'The convergences are just the zeitgeist.' What the zeitgeist explains, and what it does not. Broad worries and specific structural matches are different things, and only one leaves publishable dates.
- 'The neurodivergence angle is special pleading.' Read the PNAS Nexus paper, then say that again. What Hernández-Espinosa and colleagues actually argue about cognitive diversity as a protective factor, and what the programme carefully does not claim.
- 'The Pope did not read your book.' Correct. Nobody claimed he did. What the Babel and Jerusalem convergence actually asserts, and why influence claims would weaken it.
- 'The timeline is coincidence.' What probability arguments can and cannot show. Why this programme refuses to publish a made-up probability figure, and what the pattern actually asks the reader to weigh.
- 'Undecidability proofs prove too much.' What the Trilemma forecloses, and what it leaves open. A careful reading of Gumbau Mezquita's Unverifiability Theorem and what it does and does not license the field to do next.
- 'Why not just publish in a journal?' The economics of independent submission, and the arXiv path in progress. What non-affiliated submission actually costs, why the record does not wait, and which venues the programme is targeting.
- 'You are not peer reviewed.' Correct. Here is what the record offers instead, and its limits. An honest account of what independent convergence provides that peer review does not, and vice versa.
- 'You changed your numbers.' Yes. Publicly, with an audit trail, and it is the system working. The corrections the record has made, why they were made, and what a corpus without corrections would actually mean.
- 'You cherry-picked the convergences.' The register rules and the quarantine list. How the convergence register is constructed so that cherry-picking is structurally impossible to hide.
- 'You cite yourself.' Self-citation and self-verification are different things. What it means when the evidence chain cites its own manuscripts, and why this is not the failure mode people think.