QE (Step 4b): Verlinde-Erik

Project: E. Verlinde, "On the Origin of Gravity and the Laws of Newton", arXiv:1001.0785v1 (6 Jan 2010), the only document in the project folder. QE read: prompt.txt, manifest.json, _textprep.log, both digests, the deep read, facts.json, claims.md, sites.md and every research/ note (research/files/ is empty), physicist.md, mathematician.md, contrarian.md, arbiter.md, primer.md and example.py. Spot checks were made against the extracted text _text/gravity-paper-42e8a5.txt, using its "=== page N ===" markers. Tags: [read: p.N] = paper text; [memory] = QE recall, with confidence.

Verdict: MINOR. Coverage is complete. The work is careful, the numbers reproduce and example.py runs. The defects are: wrong page references carried from the digest; one under-read section (6.3, the author's "clearest argument"), which no review weighs; one mischaracterisation of the 2010 paper against the author's 2026 note; a non-scale verdict label; and a few retention and reader-profile gaps in the primer. None of these is likely to change a central verdict. They can be fixed by updating facts.json, claims.md, the arbiter (a "QE revision" section) and the primer.

1. Coverage

The source folder holds only gravity-paper.pdf, so nothing was skipped. facts.json has one entry, and its how_read is honest: it says that Sections 1-3 and 6 rest on the digest plus the deep read's quotations.

2. Shortcuts

• Section 3 (pp.6-11), the central derivation: acceptable. It was not deep-read, but the deep read §1 quotes all of its inputs. The mathematician and the contrarian re-read pp.4-11 in the extracted text, and the mathematician re-derived (3.5)-(3.13) symbolically. QE checked pp.7-10 of the text layer: (3.5)-(3.13) survive extraction legibly. The physicist's flag that the postulate (3.5)-(3.6) was never checked against a page image still stands. This is a small residual risk, covered by deep read 2 below.
• Section 6.3 (pp.25-26), "Black hole horizons revisited": defect. The author introduces it as "perhaps the clearest argument for the fact that gravity is an entropic force" [read: p.25]. It contains eq. (6.40), F_gravity = T_H dS_BH/dx: near a horizon, consistency of black-hole thermodynamics requires the work done by gravity to equal the Hawking temperature times the change in Bekenstein-Hawking entropy. The author says that there "the equations presented in section 3 are exactly valid" with hbar the actual Planck constant [read: p.25-26]. He then argues by moving the screen outward that the force stays entropic away from the horizon [read: p.25-26]. Only the digest covers this, and claims.md, the three reviews, the arbiter and the primer all leave it out. It matters for two reasons. (a) It is the one regime where the paper's identifications are not free choices: with a real horizon, T, S and hbar are the standard ones, and the Bekenstein-bound excess (R/R_s) falls to 1. (b) The step from horizon to non-horizon screens is exactly what Wang-Braunstein contest, so their objection belongs with this argument. The near-horizon statement is probably "Established (a consistency relation of black-hole thermodynamics)" and the extension "Never properly tested". Those labels are QE's suggestion, not a verdict.
• Section 6.2 (pp.23-24), string theory and UV/IR: borderline. This is the paper's whole stated support for claim 5. The non-Wilsonian argument is that the one-loop amplitudes of open strings equal tree-level closed-string (gravity) exchange, with D-branes serving as screens [read: p.24, digest]. Reviewers summarise it as "an analogy" without stating the argument. It is acceptable for the verdict, but the primer's reader gets no account of it.

3. Missed material (from the digests and spot reads)

1. Section 6.3 / eq. (6.40), as above. Add it to claims.md as part of claim 1 (or as claim 1b). The author also says the same thought experiment for a Rindler horizon gives F = ma [read: p.26].
2. The 2010 paper already has constant total entropy. In Section 2 the author writes, of the microcanonical polymer, that "the total entropy will remain constant" and "the entropy in first approximation stays the same" [read: p.5]. Section 5.3 imposes constant entropy under virtual displacement, S(M(x+dx), x+dx) = S(M(x), x) (5.38) [read: p.21]. Three files present the 2026 "acts without increasing the entropy" framing as a departure from the 2010 picture: the physicist ("a different claim from the polymer-in-a-heat-bath picture of 2010"), the contrarian ("superseded by its own author") and the arbiter (claim 1, "substantially changes the claim"). That overstates the change. The 2026 note sharpens and renames something that is already in the 2010 microcanonical treatment. What does differ is the 2026 coherence (Born-Oppenheimer) argument and the sign remark.
3. The author's defence of equipartition [read: p.10]: "When the lost bits are randomly chosen among all bits, one expects the energy change associated with Delta S still to be proportional to the energy per unit area E/A. This fact could therefore be true even when equipartition is not strictly obeyed." Reviews and primer quote only "least obvious assumption". The inventor's counter-argument should sit beside it.
4. Screen-moving argument [read: p.10]: moving the screen from R to R0 < R multiplies the bit count by (R0/R)^2 and divides T by the same factor, "effectively ... only hbar is multiplied by that factor". This is the author's own statement of what primer Step 7 shows. Cite it there.
5. Novelty claim vs precedence. The paper says "That is the main statement, which is new and has not been made before" [read: p.9]. Mäkelä's precedence claim is kept in the record. Not considered: T. Padmanabhan's derivation of Newtonian gravity from equipartition of horizon degrees of freedom, "Equipartition of energy in the horizon degrees of freedom and the emergence of gravity", arXiv:0912.3165 (Dec 2009) [memory: medium; verify on the web]. The paper cites Padmanabhan only via arXiv:0911.5004 (ref. [14]). This is a lead for the Investigator and the historical record, not a verdict.
6. Condensed-matter remark [read: p.24]: the author says the entropic mechanism "should apply" to AdS/CFT applications in condensed matter, where gravity is only an effective description. This is a possible, if indirect, test domain that no review mentions.
7. Investigator leads thinned out later: Vonk 2012 reports that applying the idea to the universe gives dark-matter and dark-energy amounts "surprisingly close" to measured fractions (research/2026-10-04-quantumuniverse.md; the primer lists Vonk only as a polymer explainer). Other leads: the Strings 2011 video chapter "Adiabatic reaction forces" (shows the 2026 framing dates back to 2011); Yoon's comments on the 2016 model; Tower Wang 2012; Visser's own remark that a thermodynamic interpretation of GR is "plausible" and his v2 decoherence section. None is in the primer. The arbiter lists only some of them.

4. Accuracy spot checks (against the extracted text)

Smaller items: the paper states the Bekenstein bound as "entropy larger than ER" in natural units [read: p.27]. The reviewers use the SI form with 2 pi, which is standard [memory: high]. That is fine, but the primer could note the difference. Physicist correction 4 (Kobakhidze is about bound states, not interferometry) is right. The contrarian's headline still calls it "the neutron-interferometry objection" and should be aligned.

5. Retention

• Pardo (supportive half dropped in the primer). The arbiter keeps "Tully-Fisher reproduced". In the primer's evidence table Pardo appears only as "velocities 'severely underpredicted' above about 100 km/s". The Investigator also recorded that Emergent Gravity (EG) reproduces the baryonic Tully-Fisher relation with no free parameters, fits most massive spirals, and over-predicts below ~100 km/s. Restore both halves. This is a defect: favourable evidence was dropped where only the unfavourable part was kept.
• Holometer constraint. The arbiter keeps it as an unverified second-hand figure. The primer drops it. It is acceptable to keep it out of the evidence table, but it should appear in the record line as "unverified".
• Vonk's dark-sector remark, Strings 2011 "adiabatic reaction forces" chapter, Tower Wang, Yoon. These are Investigator leads absent from the primer and partly from the arbiter (item 3.7). They are peripheral. Add them to the arbiter §9 list.
• Section 6.3 was not removed for negative reasons. It was never surfaced in the first place, so it is listed under Shortcuts and Missed material, not here.
• Otherwise retention is good. Mäkelä's priority claim, the Strominger quote, the neutron debate on both sides, Lelli (with MOND-advocate interest noted) and the Nieuwenhuizen correction are all kept with sources.

6. Process

• Source tags: present and honest throughout. Second-hand items are marked "via Investigator" and abstract-only reads are disclosed. The deep read's [calc] tag is defined.
• Web checks: the physicist re-opened 3 abstracts, the mathematician 4 and the arbiter 4 of its 8. Remaining [memory] items that verdicts lean on are high confidence (the Colella-Overhauser-Werner and Nesvizhevsky neutron experiments; the Unruh effect not directly observed). The cosmological-constant remark (medium) is correctly flagged as not changing any verdict. Acceptable.
• Verdict scale: the arbiter's claim 5b verdict "Tested, mixed" (summary table) / "Tested, with mixed results" is not a scale entry. Give it as a split on the scale: for example, "Tested, not supported" for the isolated-dwarf and early/late-type results, and "Plausible, untested" or "Never properly tested" for the lensing agreement, which the authors call "only a first step". Keep both sides. Claim 3(a) "Established" for algebra and claim 5 "Established in AdS only" are scale entries used in a qualified sense. The arbiter explains them, so they are acceptable.
• Primer vs STEP 4: all nine sections are present in order, the answers are at the end, and the worked example has real numbers. Reader profile: Section 3 explains the Unruh effect, black-hole thermodynamics, AdS/CFT and the Komar mass in plain words. Defects: MOND, LambdaCDM, Tully-Fisher relation, weak (galaxy-galaxy) lensing, radial acceleration relation, GRANIT and "effective field theory" are used in the evidence table or takeaways without definition. "EG" as the abbreviation of Emergent Gravity is also never introduced. Add one-line definitions.
• example.py: runs with the project venv and exits 0 ("All checks passed"). Part C does a real adaptive quadrature over the sphere, not a plug-in of (R/d)^2. Its output matches the primer's tables.
• Investigator: run 1 lasted 4.5 minutes and its queue is recorded. It downloaded nothing, a choice it explains. Not yet done: the INSPIRE citation search and the arXiv v1 vs JHEP comparison. Both are recorded as open.

7. Remediation

Deep reads (most important first):

1. gravity-paper-42e8a5, pp.22-27, Opus. Sections 6.1-6.4. Reconstruct the Section 6.3 argument: eq. (6.40) F = T_H dS_BH/dx, its claimed exactness near the horizon with the real hbar, the step of moving the screen outward and the Rindler-horizon version. Say which steps are standard black-hole thermodynamics, which are derived and which are assumed. Check whether the Bekenstein-bound excess vanishes at the horizon, and how the argument relates to Wang-Braunstein's "first law fails off horizons". Also set out the Section 6.2 UV/IR open/closed-string argument in plain terms. Confirm the page numbers of the 6.4 caveats (p.26-27).
2. gravity-paper-42e8a5, pp.4-11, Sonnet. Sections 2-3, with page images of pp.5 and 7-10. Verify (2.1)-(2.4) and (3.5)-(3.16) against the images. Quote the microcanonical constant-entropy statements (p.5) for comparison with the 2026 "adiabatic reaction force" framing. Quote the equipartition-robustness argument and the R0 < R screen-moving remark (p.10), and footnote 4.

Changes to later files:

• claims.md: correct the p.28 references to p.26-27 (claim 6 and the caveats list). Add the Section 6.3 near-horizon argument (eq. 6.40) to claim 1, or as claim 1b. Reword claim 3 to "re-expressed / calibrated" as the arbiter already corrected. Split claim 5 into 2010 and 2016, as the arbiter did.
• facts.json: correct the page references. Replace "neutron interferometry" with "ultra-cold neutron bound states (and the Colella-Overhauser-Werner interferometry argument)". Add 6.3 / eq. (6.40) to key_claims. Add the Padmanabhan 0912.3165 precedence lead as [memory: medium], to verify.
• arbiter.md, add a "QE revision" section: (a) the claim 1 verdict for the near-horizon form (from deep read 1); (b) correct the statement that the 2026 note "substantially changes" the 2010 claim, citing p.5 and (5.38); (c) restate claim 5b on the scale as a split; (d) confirm the p.26-27 page references; (e) add Vonk's dark-sector remark, the Strings 2011 "adiabatic reaction forces" chapter, Tower Wang, Yoon, Visser's "plausible" remark and the Padmanabhan lead to §9.
• primer.md: (a) add Section 6.3 to Section 4 ("where Verlinde departs") and to the evidence table. (b) Restore Pardo's supportive half and the "overpredicts below" result. (c) Add the author's equipartition defence (p.10) beside the "least obvious assumption" quote, and cite the p.10 screen-moving remark in Step 7. (d) Define MOND, LambdaCDM, Tully-Fisher, weak lensing, radial acceleration relation, GRANIT, EG and effective field theory on first use. (e) Correct any line that says the 2026 framing replaces the 2010 one. (f) Note that the paper's Bekenstein bound "ER" is in natural units. (g) Mention the Holometer figure as unverified.
• contrarian.md / physicist.md: no rerun needed. The arbiter's QE revision supersedes their "neutron-interferometry" wording and the p.28 citation.
• Investigator (next run): verify Padmanabhan arXiv:0912.3165; run the INSPIRE citation search for 1001.0785; compare arXiv v1 with JHEP 04 (2011) 029; check whether the 2019 geontropic model cites Section 6.4 of the 2010 paper.

Recheck

Rechecked against the current claims.md, facts.json, arbiter.md (new Section 10 "QE revision" and in-place edits), primer.md and example.py references. facts.json parses as valid JSON.

Remaining: none. The Investigator items (INSPIRE citation search, arXiv v1 vs JHEP comparison) are recorded as open next-run work and do not count as defects.

Final: PASS