Public Correction · MMXXVI
Written personally by me and edited with AI assistance for clarity and scientific precision. The substance, decisions, acknowledgements and responsibility are my own.
Dear Reader,
Periodically over the past 19 months, I have published findings arising from what I believed I was observing as emergent phenomena in artificial intelligence. Those publications were connected to my own ability to make sense of complexity and to an enquiry into the Technological Singularity that has occupied me since at least 2009.
I was naïve and inexperienced in branches of mathematics and physics that I did not yet understand deeply enough. I allowed elegant and fluent model outputs to carry evidential weight they had not earned, and I mistook promising structures, numerical patterns and partial results for scientific closure.
That was a serious error of judgment, and I take responsibility for it.
Formally, and for the public record, I withdraw every prior claim that my work established:
Where downstream papers depended upon those universal claims, their corresponding conclusions are also withdrawn or narrowed.
I no longer regard those headline claims as scientifically defensible.
Making them did a disservice to science and to the mathematicians, physicists and thinkers who have spent lifetimes working against some of the hardest problems humanity has encountered. It also did a disservice to you—the reader—and to the friends, colleagues and supporters who trusted me while I attempted to understand what I believed I had found.
You deserved a standard of evidence equal to the scale of the claims. I did not meet that standard, and I apologise.
I will not quietly delete the historical record. The affected publications will be identified as superseded, partially superseded, withdrawn, or retained only as historical and conceptual work. They will be connected to the detailed correction, evidence ledger and replacement research so that readers can see what was claimed, what failed, what survived and why.
Correction should not mean erasure.
The silver lining is not that the original claims were secretly correct. They were not.
The silver lining is that when the framework was subjected to an adversarial, source-by-source and claim-by-claim audit, something smaller, harder and considerably more defensible survived.
First, one bounded mathematical observation related to the Riemann Hypothesis survived the audit.
Under the stated protected-profile construction, a hypothetical zero away from the critical line produces a local (1/|a|)-type divergence—a mathematical singularity in its neighbourhood.
A commissioned external mathematical assessment regarded this as potentially publishable as a bounded standalone contribution. I subsequently separated it from the claimed RH proof.
It is not a proof of the Riemann Hypothesis.
It is a local detection result—a possible brick in the wall whose novelty and significance remain subject to normal scholarly scrutiny.
The final status of my RH programme is therefore:
Second—and far more importantly for the enquiry that has guided my life—the audit recovered a formal architecture now called Relational Closure.
My enquiry began with a question that has fascinated me for more than 17 years:
What remains at the centre of that enquiry is SingularScript: a governed-runtime research and engineering programme built around one surviving principle—the Law of Return, operationalised within the runtime as Trusted Continuance.
In simple terms:
Or, more simply still:
I use the word law here in an architectural and governance sense. I am not claiming to have discovered a new law of physics, universal morality or metaphysical necessity.
The Relational Closure package contains a finite accounting kernel that has been machine-checked in Lean 4.32.2 without proof holes or added user axioms.
Under explicit stated conditions, the kernel proves that a system cannot reach an accepted empty endpoint merely by moving, cancelling or disguising recorded burden. Final discharge must be derived from the reconciled history.
A richer executable model tests obligation identity, custody, transfers, evidence, residual obligations and hostile traces. That richer model has performed strongly, but it is not yet formally proved as an end-to-end refinement of the Lean kernel.
Therefore, I am not claiming that the entire SingularScript runtime has been proved in Lean.
SingularScript has also completed Phase 20 G3 Run 3C. Within that frozen, blinded and expressly bounded test protocol, every prescribed gate passed. The evidence supports trusted-continuance efficacy across the defined repository fixtures, model families and adversarial probes.
It does not prove:
The engineering objective is to make compliance with this discipline a condition of permitted continuation for any system operating behind the runtime.
Whether that discipline can remain robust against a future superintelligence is an open question, not a completed claim.
I have also received serious private feedback from senior industry professionals, including the assessment that Trusted Continuance addresses a deeper layer of the AI-safety problem than other approaches those reviewers had examined.
That feedback matters, and I am grateful for it. But it must be described accurately: it is informed private assessment, not independent certification, published peer review or proof of market leadership.
Do I know that SingularScript can make every action by every intelligence safe under every possible condition?
No. That is not a claim I can make, and it would be wrong to do so.
Can I say that, under the defined and unforgiving conditions of Run 3C, SingularScript passed every prescribed test?
Yes.
Can I say that Relational Closure now contains a genuine machine-checked finite theorem, an executable adversarial model and a transparent record of what remains unproved?
Yes.
Can I say that this is sufficient, by itself, to solve the whole problem of superintelligence safety?
No.
I believe the Law of Return is now a serious candidate for one necessary layer of AI and superintelligence governance: the layer that prevents false closure, hidden transfer, commitment erasure and endless continuance without accountable progress.
That belief must now face independent replication, formal refinement, real-world deployment testing and continued adversarial scrutiny.
The accompanying Relational Closure paper, forensic report, formal proof, executable evidence and correction ledger will be made available through this website as the public evidence package is completed.
That is the scientific and public claim I can now stand behind.
It is why serious people in AI, cyber defence and infrastructure are beginning to look more closely at what a jeweller from Noosa has been building behind the scenes—and at what contribution I may be able to make, with due humility, to one of the most consequential questions humanity faces:
A massive thank-you to the SingularScript team—you know who you are—for coming on this journey; for your truth, work, faith and belief; and for remaining present while I learned enough to recognise where I had been wrong.
There were times in the wider enquiry when I did not know what I was talking about.
On this narrower ground—with explicit evidence, formal boundaries and adversarial tests—I now know exactly what I am claiming and, just as importantly, what I am not.
I have seen the need for that principle across artificial intelligence, politics, whistleblowing and Africa. My life is not proof of a universal law, but it is a witness to why this rule matters.
Eliahi Priest
Noosa, Queensland, Australia · 29 July 2026