Work before the anchor is conceded in full and never argued against as though the programme were novel there.
AN-001 · · antecedent, conceded
Yampolskiy, R. V. (2020). On Controllability of AI. arXiv:2008.04071
the impossibility premise. Conceded in full. The estate does not claim to have been first to state that control fails.
the cannot-be-added-later requirement. Conceded in full, four years earlier than the anchor. The estate does not claim this as original.
not a concession. The field's own specialist recording that the literature on solvability was empty. Cited as evidence of the gap, never as evidence of the author's originality.
the Good Regulator Theorem and Requisite Variety, both classical, both conceded.
that external correction loses its own signal as it succeeds. Ashby, in the literature since 1956, quoted approvingly in this paper. Conceded, and it argues independently for correction inside the loop.
this is the passage that matters most and it is not a concession, it is an observation. He DERIVES the answer, states it, and discards it in the same sentence because it is not control. The estate's position is acceptance of his own derived conclusion with the demand for control dropped.
We do have some examples in which a lower intelligence manages to align interests of higher intelligence with its own. For example babies got their much more capable and intelligent parents to take care of them. It is obvious that lives of babies without parents are significantly worse than lives of those who have guardians, even with non-zero chance of child neglect. However, while the parents maybe value-aligned with babies and provide a much safer environment, it is obvious that babies are not in control, despite how it might feel sometimes to the parents. Humanity is facing a choice, do we become like babies, taken care off but not in control or do we reject having a helpful guardian but remain in charge and free.
What remains claimed: the derivation of a design requirement from the impossibility, and the epistemic rather than motivational form of the embedding. Yampolskiy's remedy at page 3 is motivational control, designing a system to desire not to cause harm, which is a preference installed from outside. The 8 December 2024 record asks for something else: 'not as rules to obey, but as truths they choose to uphold.' A preference can be optimised away by a system that rewrites its own preferences. A conclusion the system keeps re-deriving cannot.
the research record, 2026-08-08, PDF read at source
AN-002 · · antecedent, conceded
Good, I. J. (1965). Speculations Concerning the First Ultraintelligent Machine
the intelligence explosion, and control named as the open question in the same breath. NOT AN IMPOSSIBILITY CLAIM: the famous line carries a conditional, that the machine be docile enough to tell us how to keep it under control. A proviso contemplates control rather than ruling it out.
What remains claimed: unverified until read at source.
AN-003 · · antecedent, conceded
Bostrom, N. (2014). Superintelligence, chapters 8 to 13
recursive self-improvement, the control problem, motivation selection and value loading. The whole spine of the standard reference work. IMPORTANT: this book does NOT claim control is impossible. Its second half proposes control methods, capability control and motivation selection. Treating 'wrote the standard work on the control problem' as 'said control is impossible' is a category error and both lanes made it once.
What remains claimed: unverified until read at source.
AN-004 · · antecedent, conceded
Vinge, V. (1993). The Coming Technological Singularity
the post-singularity world as beyond human prediction. Whether it asserts uncontrollability as a conclusion is UNVERIFIED.
What remains claimed: unverified until read at source.
AN-005 · · antecedent, conceded
West, G. B., Brown, J. H., and Enquist, B. J. (1997). A General Model for the Origin of Allometric Scaling Laws in Biology. Science, 276(5309):122-126. doi:10.1126/science.276.5309.122
the transport-network premise. Nothing in the history of the universe has grown without bound. Every growth process that has run has terminated, and in every case the limit came from outside the growing thing: its supply ran down, its substrate was exhausted, or it destroyed the conditions of its own continuation. Not one stopped because it got better at regulating itself. The programme cites this premise and does not re-derive it.
What remains claimed: the replacement limit that operates from inside once the pipe stops governing, its number, and its dependence on what the corrector is built from.
the research record, 2026-08-12, DOI verified at source via Crossref
AN-006 · · antecedent, conceded
White, C. R., and Seymour, R. S. (2003). Mammalian basal metabolic rate is proportional to body mass to the two-thirds power. Proceedings of the National Academy of Sciences, 100(7):4046-4049. doi:10.1073/pnas.0436428100
that the value of the metabolic exponent is disputed in the literature. White and Seymour reported mammalian basal metabolic rate proportional to mass to the two-thirds power, contesting the three-quarters figure in West, Brown and Enquist 1997. Cited so the premise is never offered stripped of its contest.
What remains claimed: the programme needs the mechanism, not the number, and survives the dispute intact.
the research record, 2026-08-12, DOI verified at source via Crossref
AN-007 · · antecedent, conceded
Kozlowski, J., and Konarzewski, M. (2004). Is West, Brown and Enquist's model of allometric scaling mathematically correct and biologically relevant? Functional Ecology, 18(2):283-289. doi:10.1111/j.0269-8463.2004.00830.x
that the WBE derivation itself has been questioned in the literature. Kozlowski and Konarzewski asked in the paper's title whether the West, Brown and Enquist model is mathematically correct and biologically relevant, disputing the mathematics and the biology rather than only the number.
What remains claimed: the programme needs the mechanism, not the number.
the research record, 2026-08-12, DOI and title verified at source via Crossref and publisher landing page
AN-008 · · antecedent, conceded
Bettencourt, L. M. A., Lobo, J., Helbing, D., Kuhnert, C., and West, G. B. (2007). Growth, innovation, scaling, and the pace of life in cities. Proceedings of the National Academy of Sciences, 104(17):7301-7306. doi:10.1073/pnas.0610172104
the sublinear scaling of infrastructure (the pipes) and the superlinear scaling of information and wealth creation. Found by the same senior author (West) who set out the biological premise in 1997.
the biological-organism comparator: organisms scale sublinearly. Quoted verbatim as it appears in the abstract.
the recorded exception where the ordinary pattern breaks: cities keep going, but only when the brake is reapplied faster forever. Infrastructure is the pipes; the brake in their account is entirely external.
What remains claimed: the programme adds the internal replacement limit that this account leaves absent, and makes its number depend on what the corrector is built from.
the research record, 2026-08-12, abstract verified at source three times independently as recorded in the canonical brief for this wave; DOI verified via Crossref
AN-009 · · antecedent, conceded
Hutter, M. (2012). Can Intelligence Explode? Journal of Consciousness Studies, 19(1-2):143-166. arXiv:1202.6177
the philosophical question this programme inherits. Hutter sets out to separate speed from intelligence explosion and to consider possible bounds on intelligence, and names Chalmers 2010 as the first comprehensive philosophical treatment of the singularity in a respected philosophy journal. Hutter and Chalmers are the named owners of the question the programme continues.
What remains claimed: the replacement limit, its number, and its dependence on what the corrector is built from. When growth stops being fed through pipes, what limits it from the inside is the system's own capacity to correct itself: measurable on a running system, yielding a computable ceiling, and depending on what the corrector is built from rather than on how capable it is.
the research record, 2026-08-12, arXiv record and Journal of Consciousness Studies reference verified at source
AN-010 · · antecedent, conceded
Snell, C., Lee, J., Xu, K. and Kumar, A. (2024). Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters. arXiv:2408.03314
the comparison itself, set up as revisions in sequence against equally many parallel attempts. Conceded in full. No precedence on the direction or on the comparison is available to this programme and none is claimed.
their own qualifier on the win, carried wherever the finding is cited. The win is narrow in their words and is never reported here as more than that.
their own conclusion, that sequential revision and parallel sampling are two complementary axes for scaling test-time compute, and that a balance rather than full depth is best on the more difficult questions.
What remains claimed: Only the narrower conjunction: depth and breadth as separately scaling quantities, each with its own exponent, and the depth exponent exceeding a compute-matched breadth exponent by a margin fixed before the measurement.
read at source and hashed by the research record, 6 September 2026; the even-comparison sentence, the figure caption qualifier and the complementary-axes conclusion quoted from the paper
Prior results the record precedes
AN-011 · · precedes-result
Sharma, S. and Chopra, P. (2025). The Sequential Edge: Inverse-Entropy Voting Beats Parallel Self-Consistency at Matched Compute. arXiv:2511.02309
AN-012 · · precedes-result
Wu, Y. and colleagues (2025). arXiv:2502.10858
AN-013 · · precedes-result
Inoue, Y. and colleagues (2025). arXiv:2503.04412
AN-014 · · precedes-result
Alabdulmohsin, I. and Zhai, X. (2025). arXiv:2502.07503
AN-015 · · precedes-result
Schwethelm, K. and colleagues (2026). arXiv:2604.21106
Excluded from the convergence record, with the reason
EVR-03 · · excluded
Unidentified DeepMind publication on recursive learning
No title, authors, identifier or exact result were recorded, so the proposition map cannot be checked.
EVR-05 · · excluded
Unidentified paper on Constitutional AI limitations
No identifier was recorded. Constitutional AI's own limitations are also discussed by its authors.
EVR-06 · · excluded
Reported dissolution of an OpenAI safety team
Treating corporate conduct as evidence for a technical thesis transfers a technical term metaphorically to institutional behaviour.
EVR-11 · · excluded
Proposal described as an ARMv10 Platform Security Level specification
The cited material appears to be an unaffiliated, unsolicited mailing-list proposal. No Arm endorsement, adoption or implementation has been established.
EVR-13 · · excluded
Claimed United States patents for runtime hardware enforcement of ethical constraints
No publication numbers, inventors or priority dates were recorded, so the claim cannot be verified or compared through a claim chart.
A valuation or funding event does not test the theory. Naming is not evidence.
Retracted or withdrawn measurements
retracted estimate; not classified as a failed prospective prediction without a distinct prediction and assessment rule
recursive self-improvement growth exponent, early headline: approximately 2.24
A retracted estimate is a retracted measurement. Counting it as one failed prediction requires a distinct prediction and an assessment rule, neither of which this record supplies.
Unresolved comparisons stay visible until an owner answers them.
OC-001 · open check, owner unassigned
Does Yampolskiy's 2024 book, AI: Unexplainable, Unpredictable, Uncontrollable, publish before or after 2024-12-08?
One date decides two opposite columns. Before the anchor it is an antecedent and must be conceded like the paper. After it, it is convergence and supports the record. Neither lane has checked.
Rule until answered: cite the 2020 paper only.
OC-002 · open check, owner unassigned
Is there any source before 2024-12-08 that asserts, AS A CONCLUSION rather than a risk, condition or open problem, that control of a recursively self-improving system is impossible in principle AND derives from that a positive design requirement?
This is the actual novelty question. Yampolskiy 2020 holds the premise and explicitly declines the derivation at page 21. If nobody before the anchor made the derivation, that is what the estate owns.
Rule until answered: claim the derivation and the epistemic form only. Never claim firstness on the impossibility.
OC-T4-003 · open check, owner T2-research (scientific owner)
How do the particulars of Steunebrink, Thorisson and Schmidhuber, Growing Recursive Self-Improvers (AGI 2016, first online 25 June 2016, DOI 10.1007/978-3-319-41649-6_13) compare with the historical conjunction: education, trust and benevolent behaviour in recursively improving agents; footnote 9 on bypassing drives through hardware self-surgery; a protected motivational subsystem; continuing stakeholder testing?
A material predecessor that requires explicit treatment. It is neither a generic education paper nor shown to contain every component of the conjunction; only a scoped comparison by the scientific owner can say which components it reaches.
Rule until answered: No firstness wording for any component this paper may reach; cite the paper where the conjunction is discussed.
Attempts are counted; favourable outcomes are never counted or scored. Every proposition's own figures are in the provenance view.
ListenListen · author’s voiceListen · standard voiceResumePlayPauseThis device has no voice installed for this language, so it cannot read the page aloud.Read in EnglishListen · author’s voice (English)The author’s English voice reads aloud; the text on screen stays in your language.There is a picture here. It shows:Picturereads aloud · highlights as it goes · jump to any section