Fact-check: "Why are humans (not) solid?"

Reviewer: Claude (Opus 4.7, 1M) Date: 2026-04-23 (initial), 2026-04-25 (PDF-verified update) Draft reviewed: draft-post.md (commit-state at review time) Confidence tiers: C1 primary-verified / C2 secondary-verified / C3 inferred / C4 weak / C5 unable to verify (with PDF-update tier C1-PDF for ground-truth verification against the actual paper)

PDF-verified update (2026-04-25)

New ground truth available: Both PDFs are now in hand: - McIntyre & Fahy 2015, Cryobiology 71: 448–458 ("Aldehyde-stabilized cryopreservation"). DOI 10.1016/j.cryobiol.2015.09.003. 11 pp. - Song, LaVergne, Wróbel 2026, bioRxiv 2026.03.04.709724v1 ("Ultrastructural preservation of a whole large mammal brain with a protocol compatible with human physician-assisted death"). 17 pp. Authors are all Nectome employees; this is the Nectome 2026 preprint.

Both were extracted with pypdf (poppler-utils unavailable in sandbox). Page numbers below refer to the PDF page index.

Top corrections from the PDFs

1. The 65% w/v ethylene glycol number is correct, but the Tg figure has a subtle citation issue and the protocol does not match the 2026 preprint. McIntyre & Fahy 2015 uses 65% w/v and explicitly cites the −131 °C Tg to reference [16] (Boutron). The 2026 Nectome preprint uses 65.5% w/w ethylene glycol and 3.0–3.3% w/w glutaraldehyde, and the discussion gives the Tg as "below −130 °C for ethylene glycol:water mixtures" (page 10) — slightly different framing. Verdict: draft's "65% w/v EG, Tg ~−131 °C" is C1-PDF correct for the 2015 paper. Be careful not to conflate with the 2026 preprint, which uses w/w units and slightly different concentrations.
2. The glutaraldehyde concentration (which the draft does not give but the research note 02 §5 implies via the protocol description) is 3% w/v in McIntyre & Fahy 2015. Confirmed C1-PDF, page 4 (Table 4). The 2026 preprint uses 3.3% w/w during open-circuit fixation and 3.0% w/w in the cryoprotectant-fixative mixture. Neither paper says "tens-of-millimolar" (the draft's claim in Part 4); 3% w/v glutaraldehyde is ~300 mM, which is "hundreds of millimolar." The draft's "tens-of-millimolar" in Part 4 is a factor-of-10 underestimate. See concrete edit #5 below.
3. The "Small Mammal Prize 2016 / Large Mammal Prize 2018" attribution is NOT in the McIntyre & Fahy paper, because the paper was published in 2015 and the prizes were awarded later. The paper only acknowledges BPF as a funder ("This research was supported by 21st Century Medicine and by the Brain Preservation Foundation," page 11). No mention of any prize. The draft's hyperlink to the 2018 large-mammal-announcement page is the right primary source for the prize claim, but the paper itself cannot be cited for this. Verdict: draft's prize claim is correct; the linked source is correct; just be aware the 2015 paper itself is not the source.
4. The "processes were traceable and synapses were crisp on FIB-SEM" wording is from McIntyre & Fahy 2015 itself, not from the 2018 BPF evaluators. Direct quote, page 1 abstract: "Preservation was uniformly excellent: processes were easily traceable and synapses were crisp in both species." The draft attributes this to "the 2018 evaluators' call," which is wrong. The 2018 BPF evaluation page does corroborate the qualitative conclusion, but the exact phrase the draft is paraphrasing is the 2015 authors' self-assessment of their own rabbit and pig data. Edit needed. See concrete edit #2.
5. Nectome 2026 preprint: 14-minute window is correct; "cliff vs gradient" framing has no support in the preprint. The preprint says (page 3) "the length of the perfusability window—the time after the cardiac arrest during which blood washout needs to be initiated so that the brain ultrastructure is preserved—is about 14 min" and (page 5) "In pigs A and B, the postmortem interval was longer than the perfusability window, and the brains remained largely unfixed" (postmortem intervals 18.3 and 22.8 min). Pig E succeeded at 13.8 min postmortem interval. The discussion (page 9) says: "the perfusion needed to be initiated within about 15 min—less than 18 min, as shown by the failure of brain preservation in pig A, but possibly slightly longer than 14 min, as shown by the successful ultrastructural preservation in pig E." So the data are consistent with a relatively sharp threshold between 14–18 min, but the preprint itself does not use cliff/gradient/threshold language. The draft does not currently make this claim, so no edit needed; flagging this in case the user wanted to add such framing. The "cliff" framing should be qualified as characterized by the small-N pig data, not stated by the preprint.
6. The Nectome 2026 preprint is materially different from the 2015 ASC paper in ways the draft glosses. Specifically: (i) storage proposed at -35 °C, not -135 °C, because the preprint argues vitrified storage is "very costly" and shows that −35 °C storage gives "only about 0.1% of bonds affected at −35 °C in about 6,000 years" via non-enzymatic peptide-bond hydrolysis (page 11). This is a real protocol divergence: 2015 vitrifies at −135 °C; 2026 proposes −35 °C non-vitrified storage as a cheaper alternative. (ii) The preprint uses w/w units throughout ("All percentages in the descriptions of solutions in this section are weight per weight (w/w)," page 3) whereas the 2015 paper uses w/v. (iii) No new chemistry — same glutaraldehyde fixative, same ethylene glycol cryoprotectant. The "extension" is operational (cardiac-arrest timing window, ascending-aorta cannulation, SDS for BBB permeability, sodium azide for mitochondrial swelling — all already used in the 2015 paper) and storage-cost (proposing −35 °C). Draft does not mention this divergence; if the draft is going to cite the 2026 preprint at all, the −35 °C storage is the most newsworthy addition.

Direct quotes — McIntyre & Fahy 2015 (load-bearing)

Direct quotes — Nectome 2026 preprint (load-bearing)

Verdicts on previously-uncertain claims

New errors found that the earlier fact-check missed

1. "Tens-of-millimolar glutaraldehyde" in Part 4 is wrong by ~10×. Draft line 104: "the controlled version uses a dialdehyde at tens-of-millimolar." Actual concentration in ASC is 3% w/v = 30 g/L = ~300 mM (glutaraldehyde MW = 100.12). That's hundreds of millimolar, not tens. Fix: change "tens-of-millimolar" to "hundreds of millimolar (3% w/v ≈ 300 mM)." C1-PDF, McIntyre & Fahy Table 4, page 5.

2. "FIB-SEM" framing of the prize-winning 2018 evaluation is correct but the quote the draft uses is from 2015. Draft line 72 says: "The 2018 evaluators' call was that processes were traceable and synapses were crisp on FIB-SEM throughout the brain." The phrase is verbatim McIntyre & Fahy 2015 abstract about rabbit + pig (FIB-SEM was used in the 2015 paper too — see Fig. 8 of the 2015 paper, page 8). The 2018 BPF evaluation made an independent, similar judgment, but the draft's phrasing is the 2015 authors' self-assessment. Fix: either re-attribute ("McIntyre & Fahy reported processes traceable and synapses crisp; BPF's 2018 evaluators independently confirmed at large-mammal scale") or clearly mark the quote as 2015-paper.

3. Draft says "ramp ethylene glycol over several hours" — 2015 paper specifies 4-hour linear ramp + 1-hour plateau. Not an error, but if the user wants more specificity: page 4, §2.11, "The CPA ramp took 4 h total… we recirculated 500 mL of fresh CPA solution for 1 h." The "several hours" phrasing in the draft is fine but vague.

4. Draft does not mention the 2026 preprint at all but the research note 02 §5 cites it. This is a draft/research-note inconsistency. The research note treats the 2026 preprint as relevant (it's listed at line 105–109) but the draft only references the 2015 paper plus BPF prize pages. Either drop the 2026 reference from the research notes (if the draft isn't going to use it) or add a sentence to the draft about the 2026 extension (which would be the natural place to mention the 14-min perfusability window and the ascending-aorta cannulation challenge).

5. The Nectome preprint's BPF citation [10] is to McIntyre & Fahy 2015, not to a separate "21CM" entity. The research note 02 line 107 says "Nectome/21CM et al." but the preprint authors are all Nectome (Aurelia Song, Anna LaVergne, Borys Wróbel). 21CM is the 2015 paper's affiliation. Minor: the research note attribution "Nectome/21CM" is misleading for the 2026 preprint specifically.

6. The research note 02 line 109 cites PMC11416988 for the 14-minute number. That PMC ID is not the same as the 2026 preprint (which has DOI 10.64898/2026.03.04.709724). I haven't checked PMC11416988 directly, but based on the structure (PMC IDs that high are 2024-and-later) and the research note's own text ("Per BPF/Nectome: the 14-minute 'perfusability window'…"), it's plausible PMC11416988 is a separate Nectome paper or a 2024 BPF/PMC summary. Worth verifying that this isn't a hallucinated PMC ID; but it is not the source for the 14-min number — the 2026 preprint itself establishes the 14-min number per its own abstract.

Internal consistency between draft and research notes — what's wrong that propagated

• Draft attribution of "synapses crisp" to 2018 evaluators propagated from research note 02, line 102, which conflates the 2015 paper and the 2018 BPF evaluation: the note's quote "Preservation was uniformly excellent…" is correctly tagged to McIntyre & Fahy 2015, but the next note line cites the 2018 BPF evaluation page, and the draft collapsed both into "the 2018 evaluators."
• Research note 02 line 109's PMC11416988 citation is the only place the 14-min number appears in the draft/note ecosystem; the actual primary source is now the 2026 preprint itself, which neither the draft nor the research note quote directly. If the user wants the 14-min number in the post, the cleanest cite is Song, LaVergne, Wróbel (2026), bioRxiv 2026.03.04.709724v1, abstract.

Concrete edit suggestions (line numbers in draft-post.md)

1. Line 72: Change "The 2018 evaluators' call was that processes were traceable and synapses were crisp on FIB-SEM throughout the brain." → "In their own evaluation, McIntyre & Fahy reported 'processes were easily traceable and synapses were crisp' across both species (Cryobiology 2015, abstract); the BPF's independent 2018 large-mammal evaluation reached the same conclusion." Source: McIntyre & Fahy 2015, page 1, abstract; BPF evaluation page (already linked).

2. Line 70: The "~−131 °C" is verified. Optionally tighten to "glass transition temperature ~−131 °C (Boutron, cited as ref [16] in McIntyre & Fahy 2015)." No change strictly needed; the existing parenthetical is correct. Source: McIntyre & Fahy 2015, page 9.

3. Line 72: Optionally add a sentence about the 2026 Nectome preprint: "Nectome's March 2026 preprint extends the protocol for cardiac-arrest scenarios — defining a ~14-minute perfusability window after death within which blood washout must be initiated to preserve brain ultrastructure (Song, LaVergne, Wróbel 2026, bioRxiv 2026.03.04.709724v1)." Only add if you want to bring 2026 work into scope.

4. Line 70: Glutaraldehyde concentration is implicit in the draft. If you want to make it explicit: "…3% w/v glutaraldehyde fixative perfused for 45 minutes, followed by a 4-hour linear ramp to 65% w/v ethylene glycol…" Source: McIntyre & Fahy 2015, Tables 4–6 (page 5) and §2.11 (page 4).

5. Line 104: Change "the controlled version uses a dialdehyde at tens-of-millimolar" → "the controlled version uses a dialdehyde at hundreds of millimolar (3% w/v glutaraldehyde ≈ 300 mM)." Source: McIntyre & Fahy 2015, Table 4, page 5.

6. Optional addition near line 76 (after the shrinkage paragraph): "The 2026 Nectome preprint also reports proposing storage at −35 °C rather than −135 °C, on the grounds that non-enzymatic peptide bond hydrolysis would still affect only ~0.1% of bonds in 6,000 years at that temperature, and that vitrified storage is operationally costly (Song et al. 2026, page 11)." Only relevant if the post is going to engage with the operational economics of preservation.

Confidence on the post overall (post-PDF-review)

• Part 1 (eggs): unchanged, high. PDFs don't speak to it.
• Part 2 (body comp): unchanged. PDFs don't speak to it. The original fact-check's body-water and collagen findings stand.
• Part 3 (fixation chemistry, ASC protocol): upgraded slightly. The protocol numbers (65% EG, −135 °C, Tg −131 °C, 4-hour ramp) check out C1-PDF. The "synapses crisp" attribution needs a small re-attribution edit. The mis-attribution of the 30% shrinkage to Mikula (rather than Lu) is unchanged from the original fact-check.
• Part 4 (cell death / aldehydes): one new quantitative error. "Tens-of-millimolar" should be "hundreds of millimolar." This is a factor-of-10 error that a reader doing back-of-envelope arithmetic would spot.
• 2026 preprint coverage: draft does not engage with it. This is fine if the post is meant to be about the 2015 paper specifically, but the research note 02 §5 and the source-urls.md both list the 2026 preprint, suggesting the author meant to use it. If the author wants to use the 14-min window or the −35 °C storage proposal as material, the preprint is now PDF-verifiable and quotable.

Net: publishable with the original ~12 fixes plus #1, #2, #5 above. The 2015 paper's protocol numbers are all correctly stated in the draft. The 2026 preprint is genuinely interesting (14-min window, −35 °C storage) but the draft doesn't currently make any claims about it that need fact-checking — the relevant question is whether the author wants to add such claims.

Executive summary — top 5 issues

1. Body water figure (58%/48%) is mis-cited. The draft cites PMC10265208 for "~58% water (men) / ~48% (women)." That paper actually reports 62% (men) / 54% (women) for normal-weight adults ages 21–60. The 58/48 numbers correspond to a different subset in the same paper (and anyway the paper reports these as point estimates without 8%/6% standard deviations). This is a real error, not a rounding issue.

2. "Collagen is 25–30% of total protein mass" is contested. The specific figure the draft uses traces back to Verzár (1964) citing a 1961 review. A 2023 Scientific Reports paper (Swinehart et al., actually measured in mice) found collagen is ~12% in females and ~17% in males of total protein. The 25–30% number appears in textbooks but has weaker primary support than the draft implies. Nature cite (s41598-019-46896-0) does not actually validate the whole-body percentage — it discusses tissue-specific percentages. Recommend softening to "~20–30% by various estimates, though recent work suggests this may be overstated."

3. Penetration rate "~1 mm/hr for formaldehyde" is misleading. The cited Leica source gives Medawar's equation d = K√t with K ≈ 0.78 mm/hr^(1/2) for 10% formalin — i.e. penetration is sub-linear. 1 mm in hour 1, but only another ~0.4 mm in hour 2, and so on. The draft makes it sound linear. This is a factor-of-2 misreading at modest timescales and worse at longer ones.

4. The "2.5 second threshold" framing of Schmiedeberg 2009 is slightly off. The paper actually says proteins with residence times under 2.5 s escape capture and those above ~5.4 s are captured; 2.5 s is the lower bound, not the threshold. Also it's a study specifically of MeCP2 mutants on heterochromatin, not a general formaldehyde-crosslinking result. The draft's phrasing (~"the temporal threshold for reliable capture is ~2.5 seconds") gets it backwards — 2.5 s is the escape threshold.

5. Pubmed/37533653 cited for Mikula group's 30% shrinkage is a mis-attribution. That PMID is Lu et al. 2023 ("Preserving extracellular space for high-quality optical and ultrastructural studies of whole mammalian brains"), not Mikula. The research note correctly labels Lu et al. but the draft says "Mikula, Denk, and Pallotto's groups have demonstrated." Lu et al. do discuss ECS collapse qualitatively but I could not verify a clean "30% shrinkage" quote in the abstract — this may be from a different Mikula paper. Recommend reattributing or deleting the 30% figure.

Honorable mentions: the "40 times concentration" inversion in the acrolein literature (research notes say acrolein is "most reactive," but Shi et al. additionally say it occurs at ~40× the concentration of HNE, which tension-flips research note 03's claim that HNE/MDA are most abundant); the "Schiff base → methylene bridge" mechanism is the textbook account for formaldehyde and is broadly right there, but for glutaraldehyde it is actively debated (α,β-unsaturated oligomer Michael additions and Arg-Lys doublets are alternative mechanisms). The draft glosses this.

Overall confidence in draft: moderate-to-high on Parts 1–2 (eggs, body composition), medium on Part 3 (fixation chemistry — a few real errors), medium-high on Part 4 (the hedging is done reasonably well). Not in fundamentally bad shape, but several specific numbers and citations need attention before publication.

Claim-by-claim table

Missing counterarguments / caveats the draft should add

• Glutaraldehyde cross-linking mechanism is genuinely contested. The draft presents Schiff-base-plus-methylene-bridge as a definite account. Research note 02 footnote 24 correctly flags that the actual in-tissue species is a polymer equilibrium and that Arg-Lys doublets, aldol condensations, and α,β-unsaturated oligomer Michael additions are all proposed. A one-sentence caveat ("the exact structure of the bridges is an area of active debate; the canonical picture is likely an oversimplification") would be honest.

• Formaldehyde cross-links are partially reversible; glutaraldehyde ones aren't. The draft mentions gluta irreversibility but doesn't foreground that this is a differentiating property from formaldehyde. For readers reasoning about "why gluta specifically for ASC," that matters.

• Myelin figures are in-situ not-in-situ sensitive. The 78–81% dry weight O'Brien figure is about isolated myelin, not myelin in tissue. In situ with 40% water, wet-weight lipid drops to ~47–49%. The draft says "up to ~80% lipid by dry weight in the central nervous system" which is OK but a reader could interpret "CNS is 80% lipid" which is wrong. Research note 01 correctly distinguishes these; draft doesn't.

• Shrinkage claim is tissue-specific and protocol-dependent. The "30% volume shrinkage" is not uniform across brain regions (brain stem shows less, cortex more per Mikula work). And aging literature shows formalin shrinks brain less than paraformaldehyde+glutaraldehyde combos in some metrics. Worth a one-sentence qualifier.

• "1 mm/hr penetration" is only for immersion fixation. The draft makes this point correctly elsewhere ("this is why whole-organ fixation uses vascular perfusion"), but a reader might still come away with the wrong mental model of how fast fixatives work in the Nectome protocol. Perfusion delivers fixative to every capillary near-simultaneously, so the relevant length scale is ~10–50 μm (capillary-to-cell diffusion), not millimeters. Worth one extra sentence.

• The steelman contrary on Part 4 is present but light. The "strong form wrong / weak form right" structure is honest, but the draft doesn't actually engage with the best counterargument: most of the non-aldehyde cell-death pathways (caspase cascades, ionic dysregulation, pyroptosis, etc.) are also downstream of damage that accumulates faster than aldehyde adducts would. I.e. in many cell-death modes, aldehydes arrive after the cell is already committed to dying. Which is a tighter critique than "aldehydes are downstream." A sentence on "aldehyde accumulation is a marker of lethal stress more often than a driver" would strengthen the hedge.

• ASC doesn't preserve the synaptic vesicle contents or receptor conformations. The draft acknowledges geometry shifts but doesn't engage with the stronger criticism: even if connectome topology is preserved, molecular identity state (which receptor subtypes are where, which vesicles are filled, what's phosphorylated) is partially lost. This is a concrete gap in the case for ASC that a skeptic would raise. The draft hints at it in Part 3 ("exact geometry, active-zone nanostructure") but could be sharper.

Citation audit

Legend: ✓ resolved-and-matches / ~ resolved-but-ambiguous-or-partial / X resolved-but-doesn't-say-that / 403 paywalled/blocked / — not checked

Overall citation quality: mixed. Most big-picture facts are correct but three citations (PMC10265208, pubmed/37533653, frontiersin fcell 2023.1226044, aging-us 101450) are at minimum mis-quoted or mis-attributed. Given Claude's known ~20–50% hallucination rate on medical citations, four problem citations out of ~30 is better than the base rate, but they still need to be fixed.

Recommended fixes (concrete)

1. Line ~36 (body composition): Change "~58% water (men) / ~48% (women)" to "~62% (men) / ~54% (women) for normal-weight adults, dropping with age and adiposity." Re-verify PMC10265208 for exact numbers; if the 58/48 came from an older version of that source, cite a different primary. Currently the draft has a number that doesn't match its citation.

2. Line ~42 (collagen): Change "collagen alone is ~25–30% of all protein mass in the body" to "collagen is commonly cited as ~20–30% of body protein mass, though recent direct measurements in mice suggest the true figure may be closer to 12–17% (Swinehart et al. 2023)." Or remove the exact number and just say "the single most abundant protein in the body."

3. Line ~44 (myelin): Fine as-is, but worth clarifying "up to ~80% lipid by dry weight" → "up to ~80% lipid by dry weight (about 47–49% wet-weight, given myelin's ~40% water content in situ)."

4. Line ~63 (Schmiedeberg): Change "the temporal threshold for reliable capture is ~2.5 seconds" → "events faster than ~2.5 s residence time escape capture; events slower than ~5 s are reliably captured (Schmiedeberg et al. 2009 for MeCP2 specifically; the exact threshold is protein-dependent)."

5. Line ~64 (penetration rate): Change "~1 mm per hour for formaldehyde by passive diffusion" → "~0.8 mm in the first hour, following Medawar's diffusion law d = K√t with K ≈ 0.78 mm·h⁻¹ᐟ² for 10% formalin — so ~1.6 mm after 4 hours, not 4 mm." This is a real conceptual fix, not just a number change.

6. Line ~59 (methylene-bridge mechanism): Add a caveat sentence: "This Schiff-base-to-methylene-bridge mechanism is the textbook account for formaldehyde; for glutaraldehyde, the actual reactive species in aqueous solution is a polymeric equilibrium, and the dominant in-tissue cross-link structure is still an area of active debate (Migneault et al. 2004; Salem et al. 2010)."

7. Line ~76 (30% shrinkage / Mikula): Either (a) cite Mikula's actual Cell Reports Methods paper (Mikula et al. 2023) directly rather than PMID 37533653, or (b) drop the attribution to "Mikula, Denk, and Pallotto's groups" and just cite the shrinkage figure to Lu et al. — but verify the 30% number is actually in Lu et al. first. Current state is a mis-cite.

8. Line ~93 (acrolein levels in Alzheimer's): Fine, but tighten: "Elevated in Alzheimer's hippocampus (2× vs age-matched controls per Mohmmad Abdul et al. 2010)."

9. Line ~98 (ferroptosis pore formation): Change citation from frontiersin fcell 2023.1226044 to Cell Reports Physical Science 2023 (Shoemaker et al., "How do different lipid peroxidation mechanisms contribute to ferroptosis?") — that's the actual MD-simulation paper that shows aldehyde-only pore formation. Or JACS Au (jacsau.2c00681) for the same finding.

10. Line ~94 (MOLD/GOLD in aged tissue): The aging-us.com citation doesn't support the "accumulate in aged human tissue" claim when I fetched it. Replace with a Thornalley review or PMC6355252 ("AGEs in aging and metabolic disease") which does discuss tissue accumulation.

11. Part 4 framing (strong/weak versions): The hedging structure is good. One addition would help: acknowledge that aldehyde adducts are often markers of oxidative stress rather than drivers of the cell-death decision — i.e., in many cases the cell is already committed to dying by other pathways (ATP collapse, caspases, Ca²⁺ overload) before significant aldehyde damage accumulates. This tightens the weak form's honesty.

12. Part 3 add-on: Insert somewhere in the "what aldehyde fixation doesn't do" paragraph (~line 76) a note that molecular identity state (which receptor isoforms are where, which vesicles are filled, phosphorylation state) is partially lost even when connectivity is preserved. This is a known limitation and the draft should acknowledge it explicitly rather than letting it slip.

My confidence

• On Part 1 (eggs): high. Numbers check out; framing is fair.
• On Part 2 (body comp): medium. Body-water numbers mis-cited (62/54, not 58/48) and collagen figure is contested. Collagen reasoning is otherwise OK.
• On Part 3 (fixation chemistry): medium. Mechanism oversimplified for glutaraldehyde, penetration rate misread, temporal threshold slightly reversed, one citation (Mikula/Lu 30%) mis-attributed. Core chemistry (lysine-amine attack, gluta di-functional advantage, ASC protocol details) is correct.
• On Part 4 (endogenous aldehydes): medium-high. Cell-death framing is honest about the strong-vs-weak split. A few specific claims (MOLD/GOLD in aged tissue, ferroptosis pore formation) cite the wrong source but the facts themselves are well-supported elsewhere.
• On overall honesty: high. The epistemic-status note up-front is appropriate. The author does the "strong form wrong, weak form right" split fairly. Main risk is not over-claiming Part 4, it's the quantitative errors in Parts 2 and 3.

The post is publishable with the ~12 specific fixes above. None of the errors are so serious that the post's main argument collapses; most are factor-of-1.5 quantitative slips or citation-reattribution. I'd recommend another pass before submitting.