# Problem Frontier — bounded follow-up, 28 October 2026

The review scheduled for 20 October was missed. Hedegreen Research returned to the original nine AlphaProof Nexus Erdős result claims on 28 October. These are nine parts or variants across seven catalogue numbers, not nine completely solved catalogue problems.

This dated record adds a limited source follow-up to the July baseline. It does not replace the frozen July payload or constitute the full quarterly comparison of the original tracker. OEIS reconciliation, comparable problem additions and living-registry stock measurements were not completed. The wider scheduled tracking plan is paused; no new quarterly deadline is promised.

## What changed

- The publisher's Crossref deposit dates the AlphaProof Nexus paper's Science publication to 8 October. The version of record and supplement were not directly accessible during this review. Publication metadata is not an audit of every individual proof or of priority.
- An external replay published on 20 October reports acceptance through Lean and Nanoda for the nine published APN Erdős files and the separately corrected upper-density proof of 741(i). That makes ten check targets, including both interpretations of 741(i), rather than ten new mathematical results. Hedegreen Research matched the nine cohort solution digests, control-code hashes and successful GitHub run/job receipts. No local Lean/Nanoda rebuild was performed.
- Formal Conjectures changed its main 741(i) statement to upper density on 7 August. The old APN bundle still maps that name to a natural-density counterexample; the corrected upper-density proof is a separate April artifact.
- New work around 138 goes beyond the old consecutive-difference result. An August preprint addresses W(k)/2^k; OpenAI's October repository release contains a stronger root-growth result and Lean material. Formal Conjectures marks root growth solved, while the captured community database still labels the parent problem open. The consecutive-quotient question remains open in Formal Conjectures.
- The community database added a Lean qualification to finite problem 152's status, with an update field of 23 August. This is a registry change, not a new result by the original agent.

## What did not become established

The checked sources yielded no newly located withdrawal or refutation of the nine precisely scoped results. This is a bounded search finding, not an exhaustive mathematical or priority review. The historical eight provisional frontier results have not become eight newly certified discoveries. Result 846 remains excluded from the strict novelty count because a 2024 theorem already implied its conclusion.

All nine cohort Lean artifacts and nine accompanying prose PDFs match the July hashes. The earlier documented prose and link defects therefore remain properties of those unchanged artifacts. This does not assert that an unread final Science supplement contains the same defects.

The general remaining part of 12, upper-density question at 125 and lower-density variant of 741 remain open in the inspected Formal Conjectures files. Historical attribution of the precise 26 variant remains incompletely checked. Direct access to the seven erdosproblems.com pages returned HTTP 403; repository and Plasma assessments are attributed alternatives.

## Inspect the record

The accompanying JSON contains the nine case rows, dated source pins, comparison limits and explicit unmeasured categories. The CSV is the same case ledger in tabular form. The source-receipts JSON records captured public URLs and hashes, including access failures; it is a receipt list, not a claim that the full sources are hosted here.

Primary sources:

- https://api.crossref.org/works/10.1126/science.aej2213
- https://github.com/google-deepmind/alphaproof-nexus-results/issues/4
- https://carlostoledo.co/reports/alphaproof-nexus.html
- https://github.com/carlostoledo1891/cert-machine/actions/runs/39415085701
- https://github.com/google-deepmind/formal-conjectures/tree/5a2fb85045c86e4a184b15ce7531ba199a3c4580/FormalConjectures/ErdosProblems
- https://github.com/teorth/erdosproblems/tree/8c9d2224995163229c5ae61b642eb5cfb4a1f75e
- https://arxiv.org/abs/2608.20824v1
- https://github.com/openai/math/tree/adc7f1241b42e322a6451854ab7e4b4c146bf78a

Proof state remains `none` and mathematical outcome remains `unsettled`. The Objects in Time record establishes chronology and exact artifact identity, not mathematical certification. The original scheduled date remains visible as history; the appended events record the missed date and late, limited follow-up.
