---

Subject and independence

Role: independent reviewer in a fresh context, commissioned for refutation and given only the assignment. The reviewer took no part in writing the page, any page in its folder, the library card or the imported lemma pages, and had not seen the page before this review. The review is a focused read of the frozen text; it assigns no tier.

Frozen subject: path wiki/research/erdos_501/glazer_theorem_3_2_reconstruction.md as it stood at 2026-09-28T05:03:27Z, the page Glazer Theorem 3.2, read in full as of that time.

Artifact: E. Glazer, draft rev10, the folder-name PDF held by the library card (eight pages; the printed page numbers coincide with the physical pages). Physical pp. 3–4, the whole of Section 3 (coding conventions, Definition 3.1, Theorem 3.2 and its proof, displays (3.1)–(3.10)), were read clause by clause in the text layer and on page images; p. 2 (Lemma 2.1) and the top of p. 3 (Lemma 2.2) were read the same way for the imported lemmas; pp. 1 and 5–8 were skimmed in the text layer only, for the definition of $\mathrm{Free}_\omega$ and display (1.1) on p. 1 and to confirm that no later page restates Theorem 3.2. Page images rendered: physical pp. 2, 3 and 4. No canonical conversion sits beside the PDF.

Allowed material read: the two input pages Lemma 2.1 and Lemma 2.2 as of the same time, read in full because the whole file is what a revision read prints; their Definitions and Statement sections are what the checks use, and their proofs were not relied on; the provenance paragraph of the library card; the Statement paragraph of the problem page wiki/problems/set_theory/E0501/_index.md; docs/verification.md "Whole-claim report" and "Audit checklist" (the general list of canonical failure modes and the Erdos-specific section of the same name); docs/evidence.md "Source fidelity"; and docs/math_authoring.md in full.

Exposures: (1) the whole library card index reached the reviewer, because the revision read prints the entire file: its Overview, Read status, Companion formalization and Relation to E501 sections, including a read-status sentence that calls the folder's reconstructions author-recorded and a paragraph on the catalog's acceptance of the result; none of it was used, and every check below rests on the PDF. (2) The Status, Source, References and Formalization paragraphs of the problem page were printed with its Statement paragraph, because the page has no Statement heading to cut at; the status text was not used. (3) docs/verification.md "Durable reports and current standing" was printed with the two commissioned sections. (4) A directory listing showed the file names of two other review records in the same evidence/verify/ folder; neither was opened.

Restatement

Conventions. $\lambda$ is Lebesgue measure on $\mathbb R$. A family is any indexed family $\mathcal A=(A_y)_{y\in\mathbb R}$ of subsets of $\mathbb R$, with no measurability, boundedness or measure hypothesis on the sets $A_y$. $\mathrm{Free}_\omega(\mathcal A)$ means: there is an infinite $X\subseteq\mathbb R$ such that $x\notin A_y$ for all $x,y\in X$ with $x\neq y$, in both orders of every pair. A coding space is a standard Borel space $\mathcal O$ with a map $c\mapsto U(c)$ onto the open subsets of $\mathbb R$ such that $\{(x,c):x\in U(c)\}$ is Borel in $\mathbb R\times\mathcal O$ and $c\mapsto\lambda(U(c))\in[0,\infty]$ is Borel; the source fixes such a "standard coding" without exhibiting one, and the page exhibits one. $I_m=[m,m+1)$ for $m\in\mathbb Z$.

A profile certificate for $\mathcal A$ is a tuple $((\Omega,\nu),Z,\langle x_m,c_m:m\in\mathbb Z\rangle)$ such that: (P1) $(\Omega,\nu)$ is a standard Borel probability space and $Z\subseteq\Omega$ has $\nu$-outer measure one, the outer measure being the infimum of $\nu(B)$ over Borel $B\supseteq Z$, while $Z$ itself need not be measurable; (P2) for every $m$, $x_m\colon\Omega\to I_m$ is Borel with $\nu(x_m^{-1}(B))=\lambda(B\cap I_m)$ for every Borel $B\subseteq\mathbb R$; (P3) for every $m$, $c_m\colon\Omega\to\mathcal O$ is Borel with $\lambda(U(c_m(z)))<1$ for every $z\in\Omega$, not only for $z\in Z$; (P4) for every $z\in Z$ and every $m\in\mathbb Z$, $A_{x_m(z)}\subseteq U(c_m(z))$. $\mathrm{Prof}(\mathcal A)$ says that such a certificate exists.

The result (Theorem 3.2, display (3.5)): ZFC proves that for every family $\mathcal A$, $\mathrm{Prof}(\mathcal A)$ implies $\mathrm{Free}_\omega(\mathcal A)$. The source and the page write the family quantifier outside the turnstile; the theorem is the single ZFC sentence

$$\forall\mathcal A\,\bigl(\mathrm{Prof}(\mathcal A)\to \mathrm{Free}_\omega(\mathcal A)\bigr),$$

and nothing in the proof depends on the family being definable. The proof never uses measurability of the relation $x\in A_y$; it uses (P4) only at the countably many selected profiles, which lie in $Z$.

Checklist

• Quantifiers and scope: pass. (P3) is used for every $z\in\Omega$, as stated, to bound every column $E^s$, $s\in S$; (P4) is used only at $z_i,z_j\in Z$; the conclusion gives both $y_j\notin A_{y_i}$ and $y_i\notin A_{y_j}$ for $i<j$, which covers every ordered pair of distinct elements; the selected reals are pairwise distinct by the fiber removal, so the set is infinite. No exceptional set is dropped.
• Circularity: pass. Neither $\mathrm{Free}_\omega$ nor $\mathrm{Prof}$ is assumed in the proof; the recursion is $\omega$-long by construction and presupposes no termination; its invariant ($C_j$ Borel of infinite measure, $t_j\in C_j$, $z_j\in Z$) holds at $j=0$ and is carried by Lemmas 2.1 and 2.2.
• Model and convention changes: pass. The family is replaced by the open envelopes $V(t)$ only through (P4), the transfer the certificate itself supplies, and only at profiles in $Z$. The coding convention is stated with the two Borel properties the source requires; the page's instance (the Cantor space with a rational-interval enumeration) is labeled a compilation fill and satisfies both properties (re-derived under Weakest steps). The product measure and Borel structure on $S=\mathbb Z\times\Omega$ are the actual objects Lemmas 2.1–2.2 need.
• Finite and statistical overreach: inapplicable. No finite case, average or heuristic enters; the only counting argument is inside the imported Lemma 2.1, a proof by Tonelli's theorem.
• Uniformity: pass. The column bound $K=1$ is uniform over all $s\in S$ because (P3) holds for every $z\in\Omega$; the interchange of the sum over $m$ with the measure is countable additivity over the partition $(I_m)$; $\sigma$-finiteness of $\mu$ is exhibited by the pieces $\{m\}\times\Omega$ of measure one.
• Extremal conclusions: inapplicable, except that the page's instance writes $\lambda(U(c))$ as a supremum to show it Borel; the supremum is continuity from below over the finite partial unions and is checked under Weakest steps.
• Consequences and composition: pass. Every "so" on the page was re-derived separately: Borelness of $E$ and $x$; the column bound; null fibers; a positive section of $Q(C_j)$; the choice of $z_j$ by the meeting property; the application of Lemma 2.2; the three omissions and the two directions of independence. Lemmas 2.1 and 2.2 are consumed at exactly their stated strength (measurable set of positive measure; measurable set of infinite measure).
• Computation: inapplicable. The page carries no computation.
• Reproduction: inapplicable. The page states no rerun commands and carries no evidence folder.
• Source and verdict fidelity: faithful with one correction. The statement, Definition 3.1, the locators (Section 3, physical pp. 3–4, Definition 3.1, Theorem 3.2, displays (3.5)–(3.10)) and the proof match the PDF clause by clause. One remark alters the source's sentence about the use of the outer-measure hypothesis into a false statement about the whole of (P1) (finding F1). The standing sentences claim author-recorded status only.

Weakest steps

Measurability of the graph. The page needs $E\in\Sigma\otimes\Sigma$ so that Lemmas 2.1 and 2.2 apply, where $\Sigma$ is the Borel $\sigma$-algebra of $S=\mathbb Z\times\Omega$. Re-derivation: on the Borel piece $\{m\}\times\Omega\times\{n\}\times\Omega$ of $S^2$, the map $(t,s)\mapsto(x(t),c(s))=(x_m(z),c_n(w))$ is Borel into $\mathbb R\times\mathcal O$ because $x_m$ and $c_n$ are Borel; there are countably many pieces, so the map is Borel on $S^2$. $E$ is the preimage of the Borel relation $\{(x,c):x\in U(c)\}$, so $E$ is Borel in $S^2$. Since $S$ is a countable disjoint union of copies of the standard Borel space $\Omega$, it is standard Borel, and the Borel $\sigma$-algebra of $S^2$ equals $\Sigma\otimes\Sigma$ because a standard Borel space carries a second-countable Polish topology generating its Borel sets. The same route gives $x$ Borel and every section $E_t$, $E^s$ in $\Sigma$, as the lemmas' setting requires. This step composes with the rest by making Lemma 2.1's Tonelli argument available; no completion of $\mu$ is needed, since Tonelli holds on the product $\sigma$-algebra. The page's instance of the coding also checks: the relation $x\in U(c)$ is the union over $n$ of $J_n\times\{c:c(n)=1\}$, which is open; $U(c)$ is the increasing union of the finite unions $U_N(c)=\bigcup\{J_n:n<N,\,c(n)=1\}$, so $\lambda(U(c))=\sup_N\lambda(U_N(c))$ by continuity from below, and each $c\mapsto\lambda(U_N(c))$ depends on the first $N$ bits of $c$ only, hence is locally constant, so the supremum is Borel into $[0,\infty]$.

The column bound and the hypotheses of the lemmas. For $s=(n,w)$, $E^s=\{t:(t,s)\in E\}=\{t:x(t)\in V(s)\}$, and $t=(m,z)$ lies in it exactly when $x_m(z)\in V(s)$, so $E^s$ is the disjoint union over $m$ of $\{m\}\times x_m^{-1}(V(s))$. The product of counting measure and $\nu$ gives $\mu(E^s)=\sum_m\nu(x_m^{-1}(V(s)))$. $V(s)=U(c_n(w))$ is open, hence Borel, so (P2) gives each term as $\lambda(V(s)\cap I_m)$; the $I_m$ partition $\mathbb R$, so the sum is $\lambda(V(s))$, and (P3) at $w\in\Omega$ gives $\lambda(V(s))<1$. So $\mu(E^s)\le 1$ for every $s$, which is (2.1) with $K=1$. The remaining hypotheses: $\mu$ is $\sigma$-finite and $\mu(S)=\sum_m\nu(\Omega)=\infty$; $x$ is Borel; for $a\in\mathbb R$ with $a\in I_{m_a}$, the fiber $\{t:x(t)=a\}$ is $\{m_a\}\times x_{m_a}^{-1}(\{a\})$ because $x_m$ takes values in $I_m$ and the $I_m$ are disjoint, and (P2) gives it measure $\lambda(\{a\}\cap I_{m_a})=0$. These compose into the recursion: with $C_0=S$ Borel of infinite measure, Lemma 2.1 gives $Q(C_j)$ Borel of positive measure, and Lemma 2.2 at $t_j\in Q(C_j)$ gives $C_{j+1}$ Borel of infinite measure, so the invariant carries. As an independent check of Lemma 2.2's use: $C_{j+1}=(C_j\setminus E_{t_j})\setminus N$ with $N=E^{t_j}\cup\{s:x(s)=x(t_j)\}$; $\mu(C_j\setminus E_{t_j})=\infty$ by $t_j\in Q(C_j)$, and $\mu(N)\le 1+0$, so subadditivity forces $\mu(C_{j+1})=\infty$.

Selection in $Z$ and the two directions of independence. Since $\mu(Q(C_j))=\sum_m\nu(\{z:(m,z)\in Q(C_j)\})>0$, some $m_j$ has a section $H_j$ of positive $\nu$-measure; $H_j$ is Borel because $Q(C_j)$ is Borel and $\{m_j\}\times\Omega$ is a Borel piece. The meeting property: if $\nu^*(Z)=1$ and $H$ is Borel with $\nu(H)>0$ and $H\cap Z=\varnothing$, then $\Omega\setminus H$ is a Borel superset of $Z$ with $\nu(\Omega\setminus H)=1-\nu(H)<1$, against the infimum being one; so $Z\cap H_j\neq\varnothing$ and $z_j$ exists, with $t_j=(m_j,z_j)\in Q(C_j)\subseteq C_j$. For $i<j$, the pools decrease, so $t_j\in C_j\subseteq C_{i+1}$ and $t_j$ avoids the three sets removed at stage $i$. With the source's direction convention, $(t,s)\in E$ iff $x(t)\in V(s)$: $t_j\notin E^{t_i}=\{t:x(t)\in V(t_i)\}$ gives $y_j\notin V(t_i)$; $t_j\notin E_{t_i}=\{s:x(t_i)\in V(s)\}$ gives $y_i\notin V(t_j)$; $t_j\notin\{s:x(s)=x(t_i)\}$ gives $y_j\neq y_i$. Because $z_i,z_j\in Z$, (P4) gives $A_{y_i}\subseteq V(t_i)$ and $A_{y_j}\subseteq V(t_j)$, hence $y_j\notin A_{y_i}$ and $y_i\notin A_{y_j}$. Every unordered pair $\{y_i,y_j\}$ with $i<j$ is covered in both orders, and the $y_j$ are pairwise distinct, so $\{y_j:j<\omega\}$ is an infinite independent set.

Strongest attack

The strongest attempt was to break independence in one direction by misreading the graph's orientation. Removing only the column $E^{t_i}$ (the points forbidden by the envelope of $t_i$) would give $y_j\notin V(t_i)\supseteq A_{y_i}$ for $j>i$ but nothing about $y_i\notin A_{y_j}$, and a family whose sets $A_y$ are chosen adversarially could put $y_i$ into $A_{y_j}$ for a later $j$. The proof survives because the row $E_{t_i}=\{s:x(t_i)\in V(s)\}$ is removed as well, so every later $t_j$ has an envelope $V(t_j)$ missing $y_i$, and (P4) at $z_j\in Z$ places $A_{y_j}$ inside that envelope. The two removals are exactly (3.10) in the source, and the page's three bullets state them with the correct sections. A second attack, that removing the row $E_{t_i}$ might empty the pool, fails because $t_i$ is chosen inside $Q(C_i)$, whose definition is that $C_i\setminus E_{t_i}$ keeps infinite measure; the column and the fiber cost at most measure one. A third attack, that a positive Borel set $H_j$ might miss $Z$, fails by the outer-measure argument re-derived above. A fourth, that Lemma 2.1 might need a completed product measure or a row bound, fails: Lemma 2.1 as imported asks only for a column bound and $\Sigma\otimes\Sigma$-measurability, both established. No defect was found in the mathematics; the one defect found is a remark about which hypotheses are used (F1).

Premises

• Lemma 2.1 (positive-measure selection), source p. 2, held PDF read clause by clause; imported through the author-recorded reconstruction page glazer_lemma_2_1_reconstruction.md as of the same time, whose Statement was compared with the PDF and matches. Interface: for a $\sigma$-finite $(S,\Sigma,\mu)$ with $\mu(S)=\infty$, a $\Sigma\otimes\Sigma$-measurable $E\subseteq S^2$ with $\mu(E^s)\le K<\infty$ for every $s$, and measurable $C$ with $\mu(C)=\infty$, the set $Q(C)=\{t\in C:\mu(C\setminus E_t)=\infty\}$ is measurable with $\mu(Q(C))>0$. Applied with $K=1$; every hypothesis is established on the page.
• Lemma 2.2 (preservation step), source p. 3, same reading and the same kind of import through glazer_lemma_2_2_reconstruction.md. Interface: under the hypotheses of Lemma 2.1, for measurable $x\colon S\to\mathbb R$ with every fiber null, measurable $C$ of infinite measure and $t\in Q(C)$, the set $C\setminus(E_t\cup E^t\cup\{s:x(s)=x(t)\})$ is measurable of infinite measure. Applied at each stage; every hypothesis is established on the page.
• Both lemma pages record author-recorded standing only; the page under review names them as reconstructions and inherits that standing.
• Textbook facts used without a held source: a countable disjoint union of standard Borel spaces is standard Borel; the Borel $\sigma$-algebra of a product of two standard Borel spaces is the product $\sigma$-algebra; sections of product-measurable sets are measurable; countable additivity and continuity from below of Lebesgue measure; the product of counting measure on $\mathbb Z$ with a finite Borel measure.
• Explicit assumptions: the coding convention as stated by the source (two Borel properties), with the page's Cantor-space instance labeled as a fill; the outer measure $\nu^*$ read as the infimum over Borel supersets, a reading the source does not spell out; the family quantifier read as internal to the ZFC sentence.

Findings

F1. Severity: required. Location: "This meeting property is the only use of (P1) below." Defect: (P1) on the page comprises two clauses, that $(\Omega,\nu)$ is a standard Borel probability space and that $\nu^*(Z)=1$, and the page's own proof uses the first clause twice, for "$S$ is a standard Borel space" (to identify the Borel sets of $S^2$ with $\Sigma\otimes\Sigma$) and for "each $\{m\}\times\Omega$ has measure one" (to make $\mu$ $\sigma$-finite); the sentence is therefore false as written. Witness: the source, p. 3, after Definition 3.1, says "This is the only largeness property of $Z$ used below; $Z$ need not be measurable", a statement about the largeness of $Z$ only. Replacement text: "This meeting property is the only largeness property of $Z$ used below; the standard Borel and probability clauses of (P1) are used separately, for the Borel structure of $S^2$ and the $\sigma$-finiteness of $\mu$."

F2. Severity: suggested. Location: "The instance of the coding space given under Definitions is a compilation fill". Defect: the page also supplies the justifications of three facts that the source asserts without proof (that $E$ is Borel, p. 3, "Define a Borel directed graph"; that (3.1) is equivalent to the meeting property, p. 3; that the Borel sets of $S^2$ form $\Sigma\otimes\Sigma$, implicit in the source's application of Lemma 2.1), and names only the coding instance as supplied. The justifications are correct; the omission is one of labeling. Replacement text, appended to the Standing paragraph: "The proofs that $E$ and $x$ are Borel, that the Borel sets of $S^2$ are the product $\sigma$-algebra, and that $\nu^*(Z)=1$ is the meeting property are supplied; the source asserts these facts without proof."

F3. Severity: note. Location: "So Lemma 2.1 and Lemma 2.2 apply with $K=1$." Lemma 2.2 also needs $x$ Borel with null fibers, which the page establishes only in the next paragraph. The order follows the source (p. 4, "Thus theorems 2.1 and 2.2 apply with $K=1$. Moreover, every fiber of $x$ is null."), and nothing is used before it is proved. Optional replacement: "So Lemma 2.1 applies with $K=1$, and Lemma 2.2 applies once the fibers of $x$ are shown null."

F4. Severity: note. Location: "a map $c\mapsto U(c)$ from $\mathcal O$ onto the open subsets of $\mathbb R$". The source (p. 3) says "coding space $\mathcal O$ for open subsets of $\mathbb R$" and does not state surjectivity; the reading is natural, the instance is onto, and Theorem 3.2 never uses surjectivity, since the certificate supplies its own codes. No change needed; "onto" could be marked as a reading.

F5. Severity: note. Location: "(P1) ... where $\nu^*(Z)=\inf\{\nu(B):B\supseteq Z\text{ Borel}\}$". The source uses $\nu^*$ without defining it (p. 3, display (3.1)); the page's definition is the standard outer measure of a Borel probability measure and agrees with the completion's outer measure. It is a supplied reading placed inside the definition and could say so.

F6. Severity: note. Location: "The two measure lemmas it uses are reconstructed in Lemma 2.1 and Lemma 2.2." The imported lemmas are named and linked, but their standing (author-recorded reconstructions, not independently reviewed) is stated only on the linked pages. A clause "both author-recorded" would make the page self-contained on this point.

Verdict

Source fidelity: faithful with corrections. The statement, Definition 3.1, the coding convention, the locators (Section 3, physical pp. 3–4, which are also the printed pages; Definition 3.1; Theorem 3.2; displays (3.5)–(3.10)) and every step of the proof match the held PDF; the one required correction (F1) is a remark that overstates which hypotheses go unused.

The argument as reconstructed: sound. Every deduction was re-derived above; the imported Lemmas 2.1 and 2.2 are applied inside their hypotheses with $K=1$, and the independence conclusion holds in both orders for every pair.

Limitations: the review checks the reconstruction against the held draft rev10 and the two imported lemma statements; it does not review the proofs of Lemmas 2.1 and 2.2, the forcing module (Theorem 5.1) that produces certificates, or the companion formalization, and it makes no claim about the paper beyond Section 3. The exposures listed above did not enter the checks.

This focused review assigns no tier and changes no status.