Fact-check: "The real bottleneck in brain preservation is not the cryoprotectant"

PDF-verified update (2026-04-25)

This section supersedes the URL-based audit below where they conflict. Four ground-truth PDFs were now read directly: Ames 1968 (full text, 17 pp.), Hall 2014 (full text, 16 pp.), El Amki 2020 (full text, 14 pp.), and the Nectome 2026 bioRxiv preprint (full text, 17 pp.) which was previously inaccessible. Confidence tiers below: C1 = direct PDF quote, C2 = PDF mentions with caveats, C5 = still unverifiable.

Headline corrections

1. The "perfectly preserved synapses" framing is an overclaim. The preprint says "Several synapses can be clearly seen, with pre-synaptic vesicles and well defined, darkly stained post-synaptic densities. Slight loosening of the myelin can be observed around some processes, which however themselves were well defined, so connectomic analysis would still be possible." (p. 7). It does not say "perfectly preserved" anywhere. Draft line 11 needs softening. (C1)

2. The "cliff, not smooth gradient" framing is unsupported by the preprint. The preprint never describes the transition as a cliff. Its actual language: "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." (p. 9). This is two data points (one success at 13.8 min PMI, one failure at 18.3 min PMI) with N=5 pigs total, of which only two pigs (D and E) actually got the full ASC protocol with cryoprotectant. The "cliff" inference is not supported by the data; the preprint itself uses the phrase "possibly slightly longer than 14 min." Draft line 11 should be re-framed. (C1)

3. Resolved: "12 vs 14 minutes." The 2026 preprint uses "about 14 min" throughout (abstract, conclusions, p. 9 discussion). It does not say 12 min anywhere. Aurelia's earlier "about 12 minutes" on LessWrong predates the preprint and was a more conservative estimate; the preprint's data (pig E succeeded at 13.8 min PMI; pig A failed at 18.3 min PMI) led the authors to settle on "about 14 min" as the operational threshold. There is no contradiction — it's a refinement based on n=5 pig experiments. The draft's "unresolved questions" section can resolve this: the preprint's number is 14, with explicit acknowledgment that the true threshold is somewhere between 14 and 18 min and may be "slightly longer than 14." (C1)

4. Hall 2014: "die in rigor" verbatim, but "irreversibly" is a slight overclaim. The PDF abstract: "this is followed by pericyte death in rigor, which may irreversibly constrict capillaries and damage the blood–brain barrier" (p. 1). The hedge "may" is in the source. Page 4: "Death of pericytes in rigor, after they have been constricted by a loss of energy supply, should produce a long-lasting increase in the resistance of the capillary bed." Page 5: "Pericyte death in rigor will produce a long-lasting decrease of capillary blood flow." The draft line 47 says "they die in the constricted state. Irreversibly." — soften to "...which may irreversibly constrict capillaries" (matching the abstract) or "in a way that produces long-lasting decreases in capillary blood flow." (C1)

5. Ames 1968 methodology has one detail wrong. The basilar artery was sectioned (cut), not "blocked" by the cuff. PDF p. 1: "the procedure... involved 2 steps: (1) sectioning the basilar artery, and (2) clamping the common carotid arteries and constricting the neck with a pneumatic cuff inflated to 350mm. Hg, while artificial respiration was maintained by a tracheostomy tube." The 350 mmHg cuff was around the neck (not specifically the carotids), and it was the cuff + carotid clamps that handled the carotids; the basilar was surgically transected. Draft line 41 says "inflated a 350 mmHg cuff around rabbit necks to block both carotids and the basilar artery" — this is misleading on mechanism. Correct: "sectioned the basilar artery and inflated a 350 mmHg neck cuff with carotid clamping to stop all cerebral inflow." (C1)

6. Ames 1968 "almost the entire cerebral hemisphere" claim is correctly attributable but the actual phrasing is sharper. PDF p. 9 (Summary): "the amount of tissue affected by the vascular obstructions also increased, reaching in some cases 95% of the brain after 15 min. of ischemia." Tables 1 and 2 show 95% white-area values were the maximum, not the median, at 15 min — there's significant variability (one rabbit at 15 min showed only 1% white area; another 23%; multiple at 91-95%). The draft's "almost the entire cerebral hemisphere fails to re-perfuse" is a fair paraphrase of the maximum/typical case but elides the variability. The Zhang 2024 review's quote (which the draft inherits via the earlier audit) compresses this further. Suggested softening: "at 15 minutes, large fractions of the cerebral hemispheres — up to 95% in some animals — fail to re-perfuse." (C1)

7. El Amki 2020: scope correction confirmed and refined. Abstract: "microvascular obstruction of ~20%–30% of capillaries in the infarct core and penumbra by neutrophils adhering to distal capillary segments" (p. 2). Discussion (p. 4): "stalling of ~20%–30% of capillaries in the distal vascular network by neutrophils limited tissue reperfusion to only ~60% of baseline." Earlier fact-check correctly flagged "core+penumbra" vs "core alone." But also: the 20-30% number is the total stalled rate, not the fraction stalled specifically by neutrophils. Figures 2D-E show ~35% capillaries stalled in core, ~15% in penumbra; of those stalls, neutrophils were the cause of ~47-67% in core and ~53-54% in penumbra (Fig. 2P-Q, p. 4). So the draft's "20-30% of capillaries in mouse ischemic core are stalled by neutrophils specifically" mis-scopes twice: wrong region (core only vs core+penumbra), and wrong attribution (total stalls vs neutrophil-specific stalls). Correct version: "~20-30% of capillaries in the infarct core and penumbra remained stalled after thrombolysis, with neutrophils causing the majority of those stalls (~75% in the core, ~60% in the penumbra after t-PA), capping tissue reperfusion at ~60% of baseline." (C1)

8. Nectome preprint critiques Sparks-style preservation directly. Discussion p. 10: "contrary to the view expressed by McKenzie et al. [36], we find it implausible that immersion fixation or partial perfusion would retain enough information to reach that goal." This is a load-bearing fact for the draft's "Sparks vs Nectome" section that the draft does not currently cite. Sparks/McKenzie position is immersion fixation alone might be enough; Nectome explicitly rejects this on physical grounds in the preprint. The draft frames this as "open empirical question"; the preprint argues it's already a settled disagreement on principle. Worth surfacing in the Sparks vs Nectome section. (C1)

9. The 60°C-12-hour stress test is NOT in the preprint. That experiment is from Aurelia's LessWrong post (which describes "60°C overnight"). The preprint does not mention any thermal stress test. Earlier fact-check item 28 already flagged "12 hours" as not in the LW post; PDF reading confirms it's also not in the preprint. Draft line 119 should say "overnight" not "12 hours." (C1 negative — claim absent from PDF)

10. Sample size is much smaller than the draft implies. Only 5 pigs total, of which only 2 (D and E) received the full ASC protocol with cryoprotectant. Pigs A and B failed (PMI too long). Pigs C and D had imperfect cannula placement and showed white-matter vacuolation. Pig E is the single demonstrative result — one pig at 13.8 min PMI showed excellent ultrastructural preservation. The draft frames the 14-min finding as established with more confidence than n=1-success-pig warrants. (C1)

Direct quotes by PDF

Ames 1968 (PDF p. 1, 2, 9)

• Methodology (p. 1): "the procedure... involved 2 steps: (1) sectioning the basilar artery, and (2) clamping the common carotid arteries and constricting the neck with a pneumatic cuff inflated to 350mm. Hg, while artificial respiration was maintained by a tracheostomy tube. Ischemia was induced for periods of 2.5, 5, 7.5, 10, and 15 min."
• Reperfusion method (p. 1): "the ascending aorta clamped. Both carotids were rapidly cannulated, and a suspension of colloidal carbon was allowed to flow into the vessels at a pressure of 120mm. Hg. The colloidal carbon was a suspension prepared especially for biological use (Pelikan Werke, Guenther Wagner, Hannover, Germany)."
• Results (Tables 1 and 2): max % white areas at 15 min ischemia: 95%, 95%, 95%, 93%, 95%, 91%, 95% (multiple rabbits), but also 23%, 25%, 1%, 6% in others — substantial variability.
• Summary (p. 9): "As the duration of ischemia was increased, the amount of tissue affected by the vascular obstructions also increased, reaching in some cases 95% of the brain after 15 min. of ischemia."
• Animal: albino rabbits 1.4–3.6 kg (62 total).

Hall 2014 (PDF p. 1, 4, 5)

• Abstract (p. 1): "In pathology, ischaemia evokes capillary constriction by pericytes. We show that this is followed by pericyte death in rigor, which may irreversibly constrict capillaries and damage the blood–brain barrier."
• Mechanism (p. 4): "within approximately 15 min, grey matter capillaries constricted at spatially restricted regions near pericytes... ischaemia led to most pericytes on capillaries dying after approximately 40 min, usually at locations at which the earlier constriction had occurred."
• Mechanism (p. 4): "Death of pericytes in rigor, after they have been constricted by a loss of energy supply, should produce a long-lasting increase in the resistance of the capillary bed."
• Conclusion (p. 5): "Pericyte death in rigor will produce a long-lasting decrease of capillary blood flow, as well as a breakdown of the (pericyte-maintained) blood–brain barrier."
• Experimental conditions: cortical brain slices, oxygen-glucose deprivation + iodoacetate + antimycin (simulated ischemia in vitro). Constriction at ~15 min; pericyte death at ~40 min. Note: this is in vitro slice, not in vivo whole-brain ischemia. This is a methodological caveat the draft does not currently mention.

El Amki 2020 (PDF p. 2, 4)

• Abstract (p. 2): "despite successful thrombolytic recanalization of the proximal MCA, cortical blood flow does not fully recover. Using in vivo two-photon imaging, we demonstrate that this is due to microvascular obstruction of ~20%–30% of capillaries in the infarct core and penumbra by neutrophils adhering to distal capillary segments."
• Discussion (p. 4): "stalling of ~20%–30% of capillaries in the distal vascular network by neutrophils limited tissue reperfusion to only ~60% of baseline."
• Figures 2P-Q (p. 4): in control (no t-PA) mice: 47% of stalls due to neutrophils in core, 53% in penumbra. In t-PA-treated mice: 67% of stalls due to neutrophils in core, 54% in penumbra.
• Total stall rates (Fig. 2D-E): ~35% capillaries stalled in core (controls), ~33% (t-PA); ~15% in penumbra. Without stroke: only 0.64% spontaneous stalls.
• Model: thrombin-injection MCA-M2 occlusion in mice; t-PA-induced thrombolysis at 30 min.

Nectome 2026 preprint (PDF p. 1, 3, 5, 7, 9, 10, 12)

• Abstract (p. 1): "we establish that 14 min is the approximate 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."
• Methods (p. 3): "female Yorkshire pigs (55–70 kg, n=5) were sedated with 5 mg/kg each of telazol, xylazine, and ketamine and then injected with 250 IU/kg of heparin. After ten minutes... the animals were injected with 3120-3900 mg pentobarbital sodium and 400-500 mg phenytoin sodium. Immediately following cardiac arrest, which happened in less than 1 min after this injection, the heart was exposed; a cannula was introduced into the ascending aorta..." — this is the PAD/MAiD-analog protocol.
• Results (p. 5): "In our preliminary experiments, we used four pigs to optimize the perfusion circuit and the surgical technique... pig A: 18.3 min; pig B: 22.8 min... pig C: 7.7 min; pig D: 8.7 min... In the final pig experiment (pig E) the postmortem interval was 13.8 min to approximate the upper limit of the perfusability window."
• Results (p. 5): "In pigs A and B, the postmortem interval was longer than the perfusability window, and the brains remained largely unfixed. In pigs C and D, postmortem interval was short, but histology revealed some areas of distorted morphology, especially in the white matter... Insufficient blood washout in these pigs was caused by imperfect placement of the cannula in the ascending aorta..."
• EM result, pig E (p. 7): "The pig E brain shows excellent ultrastructural preservation: clearly visible cells with uniformly intact clear and compact membranes, and mitochondria that appear normal, with visible cristae and no examples of exploded and vacuolated (or 'popcorned') mitochondria. No 'dark' neurons are present, indicating proper fixation and the lack of mechanical disruption to tissue. Several synapses can be clearly seen, with pre-synaptic vesicles and well defined, darkly stained post-synaptic densities. Slight loosening of the myelin can be observed around some processes, which however themselves were well defined, so connectomic analysis would still be possible."
• Discussion threshold language (p. 9): "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."
• MAiD/PAD protocol caveat (p. 12): "In our experiments, the end-of-life medication was administered intravenously, and the time from the drug administration to the cardiac arrest in our experiments was very short, less than 1 min. This is not always the case with medical aid-in-dying. Prolonged agonal phase may result in poor perfusion of the organs..."
• Conclusion (p. 12): "The protocol... is compatible with human brain preservation after physician-assisted death, provided that the blood washout is initiated less than about 14 min after cardiac arrest."
• Critique of Sparks-style (p. 10): "contrary to the view expressed by McKenzie et al., we find it implausible that immersion fixation or partial perfusion would retain enough information to reach that goal."
• Storage proposal (p. 10–11): "storage at around -35°C would be possible... non-enzymatic hydrolysis of peptide bonds [would result in] only about 0.1% of bonds affected at -35°C in about 6000 years." — they propose -35°C as the economically feasible alternative to deep cryogenic storage. No head-to-head Sparks-vs-Nectome cost comparison is in the preprint.
• Limitations (p. 11): "We used a small number of pigs, and arbitrarily chose the sex of the animals — none were male... Pigs have similar body mass to humans, but smaller brains, about 100–160 g for adult pigs vs. about 1000–1600 g for adult humans. Since the total cardiac output is similar, this means that a lower fraction of the total volume of the perfusate is received by the brain, reducing (about 10-fold) the probability of embolization per brain. More importantly, filtering effects of the carotid rete (rete mirabile) — a plexus of vessels that protects the pig cerebral circulation, absent in primates — is expected to further reduce this probability." — this is a substantial pig-vs-human caveat the draft does not surface.

Specific edit recipes

New errors / caveats the PDFs reveal that earlier fact-check missed

1. Hall 2014 was an in vitro slice experiment, not in vivo. The draft uses Hall to describe what happens during in vivo brain ischemia. The mechanism is plausible to translate, but the timing (15 min constriction, 40 min death) is from cortical slices in oxygen-glucose deprivation with iodoacetate + antimycin, not in vivo MCAO. Worth a one-clause caveat.

2. Pig vs human caveats. The preprint explicitly notes (p. 11) that pigs have carotid rete mirabile (a plexus filtering the cerebral circulation) which humans lack. Pig brains are also 10× smaller relative to body mass, reducing per-brain embolization probability ~10-fold. The "this works in pig → will work in human" inference has known biological complications the preprint flags but the draft does not.

3. Nectome's preprint sample is n=5 pigs, n=1 demonstrative success. The earlier fact-check noted "small sample" but did not have the actual numbers. Pig E alone is the published "success" data point. Pigs C and D had imperfect cannulation. Pigs A and B were preliminary failures used to estimate the upper bound. The draft frames the result more decisively than n=1 warrants.

4. MAiD/PAD timing in the preprint is artificial. Pigs went from drug to arrest in <1 min. The preprint p. 12 explicitly notes real human MAiD often has a prolonged agonal phase that may invalidate the 14-min window. This is a strong caveat the draft entirely misses.

5. "Carbon black" in Ames was specifically Pelikan Werke colloidal carbon (10% soot, 9.5% gelatin, 1.3% phenol). Not a problem with the draft, just a specificity — they tested India ink first and found it clotted (footnote 4 in Ames).

6. Ames also did "bloodless ischemia" experiments (p. 6) showing that no-reflow needs blood components present during ischemia: when blood is washed out before ischemia, carbon-black perfuses fully afterward. This is mechanistically important — it means the no-reflow phenomenon is partly a blood phenomenon, not just a vessel-wall phenomenon. The draft attributes more to pure capillary closure than the source actually claims; Ames argues for "a combination of factors including postischemic hypotension, an increase in blood viscosity, and a reduction in caliber of small vessels" (p. 8). This is a subtle reframing the draft could improve.

Updated overall confidence

• Central thesis (perfusion is the bottleneck): still strong. Preprint directly supports it.
• 14-min threshold: confirmed at face value; the n=5 pig basis with n=1 demonstrative success means it's the best current estimate, not a settled fact. Draft over-confidence is a medium issue.
• "Cliff" framing: C5 → mark as draft inference. The preprint does not frame this way. Two data points cannot establish a cliff.
• Hall 2014 mechanism: confirmed verbatim; "irreversibly" softened to "may irreversibly" matches source.
• El Amki figures: numbers right; scope wrong (core+penumbra, not core only) and attribution wrong (total stalls vs. neutrophil-specific). Medium issue.
• Ames methodology: mostly right; one detail (basilar sectioned, not blocked) should be corrected. The "almost entire hemisphere" framing is fair for the maximum-case data but elides variability.
• Verdicts on items confirmed by PDFs: items 8, 9, 10, 11 in the original audit table are all now C1 (PDF-verified). Items 1, 2 in the original table — for which the preprint was inaccessible — are now PDF-verified with corrections (see corrections #1 and #2 above).

Confidence ratings on original audit verdicts

Executive summary

1. The "14 minutes" number is right, but the "cliff, not smooth gradient" framing is an inference the cited preprint does not itself explicitly make. Secondary summaries confirm "<14 min ≈ intact" and "~18 min ≈ evident cellular damage," but no source I could reach describes the shape of the transition between them. "Less a smooth gradient, more a cliff" is the author's extrapolation from two data points; it should be flagged as such. (C2)

2. The draft is internally inconsistent about "12 vs 14 minutes." Aurelia's own LessWrong post says "about 12 minutes" with "a little wiggle room" and she has recorded 4:30 cannulation in pigs. The 2026 preprint reportedly uses 14 min as the outer bound. The draft treats 14 as the canonical number throughout without acknowledging the 12/14 discrepancy (which the author herself flags as an open question in the final section, so this is partially handled).

3. Two misattributed citations. (a) The brain-capillary-diameter paper is Sargent, Bonney, Shih et al. 2023, not "Chakraborty et al." (PMC10503221 is correct, but the author-string in the draft is wrong.) (b) Research note 01 cites PubMed 835156 as "Little, Kerr, Sundt 1977"; that PMID is actually Fischer, Ames, Hedley-Whyte, O'Gorman 1977. Draft inherits neither citation directly, but the Sargent mis-citation appears in the post and should be fixed.

4. The "Tomorrow Bio 2026" quote chain is partially sourced to the wrong URL. The exact quote "Since 2024, we can not improve much anymore based on CT scan, as close to 100% of the brain shows sufficient CPA concentration for vitrification" is in the founder letter (tomorrow.bio) — not in the emk.bio 2026 roadmap page, which has only weaker language ("While we still plan to improve our CT scan results, there isn't that much to do anymore"). The draft attributes this to the emk.bio roadmap. Verification requires the founder letter; swap the citation.

5. Poiseuille framing is physically correct first-order, but under-caveats in-vivo non-Newtonian behavior. The draft acknowledges that RBC deformation matters but still presents r⁴ as essentially the governing law. Capillary flow is Fåhræus-Lindqvist-dominated, plug-flow, non-Newtonian, with strong RBC-deformation coupling. This isn't wrong enough to break the argument (direction of effect is the same; a capillary that narrows by 2 μm from the RBC-already-deforming regime does become impassable), but a one-sentence caveat would be fair.

6. "37% diameter reduction at pericyte sites" is from coronary, not cerebral, ischemia. The O'Farrell 2017 eLife paper (PMC5705208) is cardiac. The research note correctly flags this ("cardiac models"); the draft paragraph at lines 47–52 elides it, reading as if it were direct brain data. A one-word fix ("cardiac no-reflow literature") would resolve this.

7. "El Amki 2020: 20–30% of capillaries in mouse ischemic core are stalled by neutrophils." The paper's actual language is "infarct core and penumbra" or "distal vascular network." "Ischemic core" alone is narrower than the paper claims. Minor — fix by adding "and penumbra."

8. The McKenzie et al. 2024 quote is verbatim-correct, but the venue attribution in the draft (it just says "PMC11416988") is opaque. Venue is Frontiers in Medical Technology. Adding "(McKenzie et al., Frontiers in Medical Technology 2024)" would make it citable. This was the single highest-fabrication-risk claim in the spec, and it checks out.

9. BPF 5 nm EM criterion: confirmed verbatim. "5nm pixel size" appears exactly that way in the prize rules PDF (page 3).

10. M22 composition: essentially correct, total is 9.345 M (draft says "9.3 molar" — ok, order-of-magnitude accurate). Component list matches Biostasis source exactly.

11. The "cooling skin can't cool the brain" section checks out, numbers match sources. The "5 hours diffusion time to 5 cm" estimate is dimensionally right; the Michenfelder Q10=2.3 figure is accurately cited.

12. The 400-mile capillary figure is sourced correctly to Cipolla (NBK53086 does contain that exact number); also confirmed the "every neuron has its own capillary" framing, but the specific "~40 μm" intercapillary distance figure is from the research note and not directly supported by this Cipolla URL — add a secondary source or hedge.

Overall confidence in draft thesis: high. Overall confidence in specific quantitative claims as stated: medium. Nothing is catastrophically wrong; the pattern is over-precise framing of things that are actually ranges or extrapolations.

Claim-by-claim audit table

Steelman of the vitrification-first position (what Alcor / Tomorrow Bio would reply)

The draft's central thesis is: "perfusion of capillaries, not cryoprotectant chemistry, is the bottleneck; therefore ASC's fix-first architecture wins." A steelman of the vitrification-only rebuttal, which the draft treats fairly but somewhat briefly:

1. ASC is irreversible in a way vitrification arguably isn't. Glutaraldehyde cross-links proteins permanently. Even under heroic future technology, revival from a cross-linked brain requires whole-brain emulation / upload — you cannot biologically revive the preserved tissue. Vitrification preserves tissue in a state that in principle is biologically reversible (rewarming a vitrified kidney has been done; rewarming a vitrified human brain has not, but it isn't physically ruled out by chemistry). The draft treats "revival" as if it's equally hard both routes. Alcor would say ASC has already made the commitment to the harder-seeming route.

2. The "perfusion problem" applies to ASC too. ASC also requires perfusion: the glutaraldehyde has to get into the capillaries. If capillaries close at 14 minutes for CPA, they also close at 14 minutes for glutaraldehyde. The draft implicitly acknowledges this ("ASC does a cryoprotectant ramp too") but the central thesis — "perfusion is the bottleneck and ASC solves it by fixing early" — doesn't fully explain why fixing early helps with the already-closed capillaries. What it does is stabilize capillaries that were still open when you started, so that they survive the later CPA ramp. That's a real benefit, but it's a benefit at the margin, not a solution to the no-reflow problem itself. If you're 15 minutes post-arrest, ASC has the same unreachable-tissue problem vitrification does.

3. Sample-size asymmetry. Nectome's publicly audited sample is very small. A small number of best-case pig runs vs. hundreds of human Alcor cases is not yet a methodology comparison. Alcor's critique would be: "Show us the distribution, not the best sample." The draft gestures at this ("no head-to-head") but the direction of the remaining uncertainty cuts both ways — Nectome's rigorous best case may or may not translate to an operational human clinical result.

4. The "audit regime" argument can be flipped. Tomorrow Bio is introducing EM sampling in 2026; Alcor hasn't made that public update. But (a) EM sampling destroys tissue, (b) the information gain from EM on a handful of samples may or may not dominate the information loss from destroying part of a patient, and (c) the calibration of EM results from stored human patients to model-organism experiments is itself assumption-heavy. Aurelia's "just do the rat experiment" is a strong ask; running enough rats under enough varied conditions to calibrate a production clinical protocol is not cheap. Tomorrow Bio / Alcor would reasonably say "prioritize human protocol quality over model-organism audits."

5. M22 / VM-1 toxicity is a real constraint. The draft treats cryoprotectant chemistry as solved. It isn't. Even optimal CPA held at terminal concentration causes cellular damage on the time scale of hours, and we don't fully understand the mechanism. This matters because some fraction of vitrification-era preservation-quality loss is from CPA toxicity, not from perfusion failure. The fix-first-with-glutaraldehyde approach doesn't remove this concern — it replaces one set of chemical insults with another (glutaraldehyde is also toxic, and cross-links some things you might wish remained uncross-linked).

6. "CT looks fine" is a cheap shot. The draft's framing that CT-only QC is near-useless is correct at the connectome-traceability resolution, but Alcor and Tomorrow Bio would say: CT at least verifies the brain is vitrified and not fractured, which is a necessary condition for any preservation quality. Treating CT as "epistemic trap" rather than "useful coarse filter" is slightly unfair. The draft's own framing ("CT is genuinely useful for one specific thing: did the brain vitrify") handles this, but the tone of the paragraph still reads as if CT is near-fraudulent when it's actually just insufficient.

7. The Sparks framing as presented is slightly strawmanned. "Sparks: cost is the bottleneck" and "Nectome: quality is the bottleneck" is a fair-ish dichotomy, but Sparks-style proponents would object that a century of neuroanatomy has in fact demonstrated ultrastructural preservation with aldehyde fixation alone. The draft acknowledges this at line 84 but then does not fully engage the "formaldehyde-fixation-is-already-gold-standard" literature as a steelman. The genuinely missing experiment (matched-timing Sparks-vs-Nectome head-to-head) is identified; the draft is fair on this.

Overall: the steelman is partially done but not exhaustively. The three missing beats are (1) the reversibility asymmetry of fix-first, (2) the margin-vs-core nature of ASC's no-reflow "solution," (3) CPA-toxicity being a real separate problem that fixation doesn't make go away.

Missing counterarguments / gaps the draft should acknowledge

1. Perfusion at "open" capillaries ≠ delivery to all tissue. If 20-40% of capillaries stall and those are randomly distributed at the microscale, then perfusion can deliver fixative/CPA to 60-80% of tissue through still-open capillaries — but the remaining fraction is unreachable through those same capillaries no matter what you do. The draft's "ASC solves perfusion" elides this. ASC makes perfusion through still-open channels more effective; it does not turn previously-closed channels back on. A fair version would be: "ASC locks in whatever perfusion you were able to achieve, and stabilizes vessels against further collapse during the CPA ramp, but it does not rescue tissue that is already microcirculatory-unreachable."

2. Pig ≠ human. Pig cerebral circulation is a reasonable analog but not identical; pig skulls are thicker, cerebral autoregulation is different, and the MAiD analog in pigs involves controlled anesthetic cardiac arrest that isn't clinically identical to human MAiD. The "this works in pig → will work in human" inference is the main empirical gap in the 2026 preprint. Draft glosses this (line 99-ish talks about the 2026 preprint as if it straightforwardly closes the human case).

3. Aldehyde fixation is not structurally lossless. Glutaraldehyde itself causes tissue shrinkage (5-15% typical in EM prep), crosslinks in a way that can obscure some molecular features (certain epitopes, small molecules), and is not reversible. Some connectomic information may be preserved while some molecular-state information (neurotransmitter levels, phosphorylation states) is destroyed. For the "preserved for revival" framing the draft uses, this matters. The draft does not mention it.

4. "Rat-left-out experiment" is cheap in the lab, not cheap for a clinical cryonics org. Running the rat experiment Aurelia wants requires running rats under clinically-realistic field conditions (not just lab conditions) and comparing EM quality. For an org like Alcor, the infrastructure gap between "preserve a patient in a rescue scenario" and "preserve a rat under realistic timing in a controlled environment" is real. This isn't a defense of not doing it, but the draft's characterization of the experiment as trivially cheap and low-effort undersells how many variables you'd need to control. Nectome has built this rigor; others may legitimately not yet have.

5. No discussion of warm-ischemia variability across MAiD patients. MAiD actual-practice varies: barbiturate vs curare protocols, IV vs oral, in-hospital vs at-home. The "14 minutes from cardiac arrest" window is defined cleanly in the lab but the clinical MAiD protocol affects what the 14 minutes looks like. This is the downstream translation question the draft should at least flag.

6. The "CT-only providers are epistemically bankrupt" framing is strong. A more honest version: CT-only providers have bounded confidence in the connectome-preservation question and are not claiming otherwise. The epistemic failure would be a provider who claims connectome-grade preservation based on CT. Alcor's public language is actually quite hedged on this point. The draft could be fairer here.

7. No discussion of which ultrastructural features are actually needed. The draft assumes connectome-traceability (BPF 5 nm criterion) = necessary. It might be sufficient for future revival via upload, but (a) the revival threshold isn't known, (b) some subset of molecular state may matter and isn't captured at 5 nm EM, (c) molecular-specificity light-microscopy approaches (Nature 2025, s41586-025-08985-1) may substitute for EM on this exact question. Draft takes 5 nm EM as the bar; could caveat.

Citation audit

Recommended fixes (concrete edits)

High priority (factual correction)

1. Line 29 — Fix author mis-attribution of the capillary diameter paper. Change: (Chakraborty et al. 2023) → (Sargent, Bonney, Shih et al. 2023). Optionally add title: "Endothelial structure contributes to heterogeneity in brain capillary diameter."

2. Line 29 — Caveat the "outer wall puts the whole tube at about 7-9 μm" claim. The 7–9 μm outer diameter is not directly in the Sargent 2023 excerpt. Either find a direct source for outer diameter or soften to "outer ~7-9 μm depending on wall thickness" (matches research-note language).

3. Line 49 — Fix scope of El Amki 20-30% stat. Change: in mouse ischemic core are stalled → in the mouse infarct core and penumbra are stalled.

4. Line 52 / footnote — Acknowledge the 37% figure is from coronary, not cerebral, ischemia. Change: individual capillaries at pericyte sites show ~37% diameter reduction → individual capillaries at pericyte sites show ~37% diameter reduction in coronary no-reflow models, with analogous but less precisely-quantified constriction in cerebral ischemia.

5. Line 113 — Fix Tomorrow Bio citation. The "Since 2024, Tomorrow Bio cannot improve much anymore based on CT scan alone" quote is from the founder letter, not the emk.bio roadmap. Keep both citations but attribute the specific sentence correctly. Also: the paraphrase should more closely match the source ("we can not improve much anymore based on CT scan, as close to 100% of the brain shows sufficient CPA concentration for vitrification"). Current draft's phrasing adds "alone" which isn't in the source.

6. Line 119 — "60°C for 12 hours" should be "60°C overnight" (matching LW post). The "12 hours" appears in the research note but not in the LW source I verified.

Medium priority (framing)

7. Line 11 — Soften the "cliff" framing. Suggested: This is less a smooth gradient and more a cliff. → The two reported data points — <14 min intact, ~18 min damaged — suggest a fairly sharp transition rather than a smooth gradient, though the preprint reports two data points not a full curve. This protects against the "how do you know it's not a gradient you've sampled coarsely" objection.

8. Line 29 — One-sentence non-Newtonian caveat. After the Poiseuille paragraph, add: (Capillary flow isn't strictly Poiseuille — RBC plug flow and Fåhræus-Lindqvist effects make the effective viscosity depend on vessel size — but the r⁴ intuition is the right first-order framing and what cerebral hemodynamics teaching uses.) This caveat appears in the research note (line 20) and should graduate into the post.

9. Line 82 — Acknowledge that ASC's perfusion also fails when capillaries are already closed. Add something like: (ASC's advantage is at the margin: it stabilizes still-open vessels so they survive the long CPA ramp. For tissue already unreachable through closed capillaries, ASC has the same problem vitrification does.) This addresses the "fix-first-solves-no-reflow" implication that isn't quite warranted.

10. Line 121 — Soften "no principled basis for quality claims." Suggested: A provider that doesn't invest in EM audits has no principled basis for its quality claims → A provider that doesn't invest in EM audits has no *direct* evidence for connectome-grade quality; CT-only evidence is consistent with that quality but does not establish it. This is less combative and more true.

Low priority (optional)

11. Line 35 — Add direct source for "every neuron within ~40 μm of a capillary." Cipolla confirms "every neuron has its own capillary" but the 40 μm intercapillary-distance number is from the research note, not directly verified against this URL. Either add a second citation or drop the μm number.

12. Line 60 — Add a source for "1-2% increase in brain water raises ICP hugely." The Monro-Kellie doctrine paper doesn't contain this specific figure. StatPearls ICP or a physiology textbook could back it.

13. Line 66 — Name the venue on McKenzie 2024 quote: (McKenzie et al., *Frontiers in Medical Technology* 2024).

14. Line 105 — "Alcor's CT page" — the URL is dead. Replace with cryonicsarchive CT calibration page (which is in the research note sources list and is still live).

15. Line 202 (unresolved questions) — Drop the "~90% capillary volume collapse" unresolved item. The draft already handles this well in the main text (line 52) and acknowledging it again as "unresolved" weakens the handling. Or reframe as a documented-in-literature question rather than an open one.

Steelman additions

16. Somewhere in the "Sparks vs Nectome" section, add the reversibility asymmetry point (steelman #1 above): ASC commits to the upload/emulation revival path in a way vitrification-only arguably doesn't. This is a genuine difference and the draft currently treats the two as symmetric on revival feasibility.

17. In the scan-modality section, soften the "CT is a trap" framing by explicitly acknowledging that Alcor's and Tomorrow Bio's own public communications are not over-claiming connectome-grade quality. The critique should be of the field's reliance on CT, not of its providers' honesty about CT.

Auditor confidence summary

• Draft's central thesis ("perfusion is the bottleneck, not cryoprotectant chemistry"): strong, well-supported.
• Specific quantitative claims: mostly accurate with some overstated precision.
• Citation hygiene: mostly good; two mis-attributions (Sargent vs Chakraborty; Tomorrow Bio quote URL) and two unverified-at-check-time items (Alcor CT page dead; 12-hour stress test vs overnight).
• Steelman of opponents: partially present, three missing beats.
• Audit philosophy section: largely fair but could soften "epistemic trap" framing.

If I had to summarize: the physics is right, the numbers are mostly right, the framing is occasionally sharper than the underlying evidence warrants, and the steelman of the vitrification-only position is present but could be more thorough. No "likely-hallucinated" claims found in the post itself. The highest-fabrication-risk claim (McKenzie 2024 quote + paper existence) checks out cleanly.