Skip to the content
Aletheia

Collapse, Not Computation · source record · webpage

The Erdős Unit Distance Problem and Penrose 'Understanding': What might stay out of the reach of classical computational AI?

Jhong, Eugene; Strange Loops In My Head (Substack) · 2026

Verified — held in project librarySRC-JHONG-ERDOS-2026
verification note
Held in the project library (research/). Published 21 May 2026. Consulted as the PDF in the project library. This essay is the source of the non-computability rung's structure in this case, including the "escalator, not the steps" formulation and the audit of how Penrose's concrete examples have aged.
reliability notes
  • The case editor's own essay; philosophical argument and curation rather than original research, and substantially composed of quoted AI-model outputs which the essay labels as such.
  • It quotes the critic's position verbatim (the @psychiel posts) and is the only verifiable record of that wording available to this import — the linked X post returned HTTP 403 to an unauthenticated fetch and was never read directly.
  • Notably even-handed against interest: the essay concedes that Penrose's concrete testable examples have 'mostly settled against' him, and that the examples aging best are the least testable ones.

Evidence drawn from this source

Qualifies· weakORCH-E026

The editor's own audit: Penrose's testable examples have mostly settled against him

what the source states
The essay surveys the concrete examples Penrose offered as marks of non-computable understanding and concludes that those which were testable have mostly settled against him — most cleanly the pawn-fortress chess position released by the Penrose Institute in 2017, which modern engines handle correctly, and now the Erdős unit-distance conjecture. It argues that the examples aging best are the least testable ones. Against that, it advances the distinction it calls "the escalator, not the steps": the claim was never that any particular insight is permanently non-computable, but that the open-ended capacity to keep generating new frameworks is — so each formalized insight becoming computable in hindsight is what the view predicts rather than evidence against it.
what we infer
Included as an honestly-labeled editorial source rather than as evidence of fact: it is the case editor's own essay, and its value is that it states the Penrose-side response in its strongest form while conceding the empirical scoreboard. The concession is what makes it usable — an advocacy piece that reports its own side losing the testable cases is more informative than one that does not. The essay is also the only verifiable record available to this import of the critic's wording: the linked X post returned HTTP 403 to unauthenticated fetch and was never read, so the tweet is not cited as a source anywhere in this case. Graded weak because it is argument and curation, not research.
limitations
  • The case editor's own essay — not independent of this case's editorial judgment.
  • Philosophical argument and curation rather than original research; substantially composed of quoted AI-model outputs, which the essay labels as such.
  • The 'escalator, not the steps' distinction is precisely the move that critics call unfalsifiable; the essay presents it without resolving that objection.
  • The social-media claim that prompted this line was never directly readable and is not cited.
Jhong, Eugene (2026)Verified — held in project libraryORCH-C040ORCH-C041