Last updated 2026-09-04in-progress

Active Candidates

The comparison record for SGL-001's active — every candidate scored on mechanism, human-relevant evidence, manufacturability, and regulatory tractability, framed by Signal's declared MSC secretome (exosome-rich conditioned medium) development focus. Higher-scoring alternatives are tracked as benchmarks, not adopted.

Signal's declared development focus is an intranasal MSC SECRETOME (EXOSOME-RICH CONDITIONED MEDIUM) therapeutic for cognitive function — a business decision the operator has made, not a research conclusion this document reaches. agents/formulation.md sets four axes every candidate is scored on regardless of focus: mechanism (defined or definable, and what it is), human-relevant evidence, manufacturability, and regulatory tractability. This document leads with the declared focus below, then holds every candidate — the focus candidate included — to the same honest scoring. A higher-scoring alternative is recorded as a benchmark, not adopted.

Scope of this run. The prior run (2026-08-28) covered three columns only — human intranasal precedent, regulatory class, molecular size/stability — and deliberately withheld a verdict. The following run (2026-09-01) added the four-axis scoring below and, before this framing update, named a per-axis "current lead" independent of any declared focus. This run doesn't touch that scoring table — the facts and scores are unchanged — it only corrects how the artifact frames them, per the operator's scope decision. The presentation/device decision (presentation-and-device.md) still doesn't exist, so cost/supply and combinability stay open. The excipient system (formulation-design.md) was created on 2026-09-03 — see Not assessed this run, below, for what it does and does not settle for this comparison.

The comparison itself stays active-agnostic, even though the product's focus is not. Extracellular vesicles are one candidate among four classes — characterised exosomes, MSC secretome/conditioned medium, defined peptides, and combinations — scored the same as every other candidate here: on its evidence, not its incumbency or its declared status. The declared focus changes what Signal is building; it does not change what this table finds.

Development focus

Signal's current development focus is the MSC secretome (exosome-rich conditioned medium) candidate; higher-scoring alternatives are tracked as benchmarks, not adopted.

The intended active is the exosome fraction of the secretome. Full exosome purification is cost-prohibitive at consumer nasal-spray scale, so the product is unpurified MSC secretome that retains the exosome fraction rather than a purified isolate. Remaining secretome components — soluble growth factors and cytokines — may contribute additively; this is not established and is not relied on for the core mechanism.

This is stated as a business decision, not a finding of this comparison. The four-axis scoring below is unaffected by it and stays fully honest in both directions — including where it doesn't favour the focus candidate.

On that scoring, MSC secretome (exosome-rich conditioned medium) — the focus candidate — scores weak-to-weakest across the board: manufacturability is weak because the product is not a single defined particle population, its own release-specification problem; regulatory tractability is weak, the same secreted-product logic as EVs with no approved comparator; human-relevant evidence for a cognitive claim specifically is weak — one human intranasal trial exists and it's non-neurological (asthma), and the cognitive evidence is preclinical only. Its one relative strength is mechanism, marginally better than naive EVs' because named trophic-factor classes (e.g. IGF-1/VEGF/TGF-β1/HGF-type) are more individually characterized than EV cargo — though the product is still a variable mixture, not one testable target.

Naive/unmodified MSC- or adipose-derived exosomes (EVs) — the previous focus candidate — is now excluded. It remains, unchanged, the weakest candidate compared on three of the four axes: mechanism is undefined and undercut by the miRNA-stoichiometry problem mirna-cargo.md documents; manufacturability is the open CDMO-readiness gap cdmo-readiness.md already records; regulatory tractability is the weakest in the field, full BLA with no approved comparator. Its one real strength — human intranasal cognitive precedent (Xie et al., n=9) — is itself thin: open-label, uncontrolled, unblinded.

Benchmarks

Candidates whose four-axis scores exceed the focus candidate's, tracked here for comparison — not under consideration as a replacement for the declared focus, and not implied to be one.

Intranasal insulin scores highest across all four axes among every candidate compared:

Mechanism — defined. Insulin-receptor signalling in the CNS is a characterised, testable pharmacology, not a proposed one.
Human-relevant evidence — the deepest controlled human cognitive dataset of any candidate compared: SNIFF (NCT01767909), Phase II, ~240 participants, 12 months, randomized, placebo-controlled, in the exact population (MCI/Alzheimer's) the product targets. Real heterogeneity by sex, APOE genotype, and device is documented, not hidden — see the Comparison table below.
Manufacturability — strong. An approved biologic with decades of industrial cold-chain and fill-finish precedent.
Regulatory tractability — strong relative to every other candidate. Off-label nasal-device repurposing of an already-approved drug is a materially easier starting point than any new biologic entity, cell-derived or not.

Intranasal oxytocin also scores above the focus candidate on every axis, including human-relevant evidence — the focus candidate's cognitive-specific evidence is preclinical only, while oxytocin's is a real, if genuinely mixed, human trial base — see the Four-axis scoring table below for both.

Insulin is not a regenerative or tissue-repair active; it would not support a "nose signals the brain to repair itself" claim the way an exosome or secretome candidate would, mechanistically or narratively. That's reported as a fact about the candidate, not as a reason to prefer it — the declared focus is the MSC secretome (exosome-rich conditioned medium) candidate regardless. This scoring gap is reported plainly, per the brief's rule that a declared focus must never soften a comparison finding, and stays recorded here as a benchmark rather than a recommendation, which is not this document's call to make.

Comparison

Active	Class	Human intranasal precedent	Regulatory class (US)	Molecular size and stability profile
Naive/unmodified MSC- or adipose-derived exosomes (EVs)	Extracellular vesicle	Three registered human trials found, all Phase I/II, none placebo-controlled at meaningful scale except one: Xie et al. 2023 / NCT04388982 (adipose MSC exosomes, mild-moderate Alzheimer's, n=9, open-label, dose-escalation 2/4/8×10⁸ particles, medium dose showed the best cognitive signal); NCT06598202 (umbilical-cord MSC exosomes, ALS, n=20, randomized double-blind placebo-controlled — the only controlled human intranasal EV trial found, still recruiting); NCT07105371 / Prodromos 2025 (allogeneic UC-MSC exosomes, n=18 completed, ALS/Kennedy disease/congenital myasthenic syndrome/Lewy body dementia, no adverse events, "near improvement" reported in most). Only the Xie trial is a cognitive/dementia indication — the other two are motor-neuron and movement disorders, a different mechanism-of-action claim than Signal's own "cognitive clarity, focus, memory" framing.	No approved product in any indication (see `jurisdiction-map.md`). FDA treats exosomes as excluded from HCT/P status (21 CFR 1271.3(d)(3)) and regulated as a drug/biologic under PHS Act §351 — full BLA pathway, no shortcut identified.	30–160 nm particle population (per Xie et al.'s own NTA data); freeze-thaw is destructive (23–36% particle loss per cycle, Ahmadian et al. 2024, already cited in `sgl-001-spec.md`); −80°C is the consensus long-term storage condition; aggregation at high concentration is a known formulation risk. This is the most stability-fragile candidate on this list by a wide margin.
Engineered or cargo-loaded EVs (e.g., miRNA-, BDNF-, or drug-loaded exosomes)	Extracellular vesicle, modified	None found. Every engineered/loaded intranasal EV study located this run is a rodent or in vitro model (e.g., intranasal miR-206-3p antagomir-loaded EVs in an Alzheimer's mouse model). No human intranasal trial of an engineered or loaded EV was found in this search.	Same base regulatory class as naive EVs, plus an additional layer: a loaded cargo (nucleic acid, small molecule, or protein) each carries its own characterization, potency, and safety burden on top of the vesicle's. No precedent exists for how FDA would treat the combination product question specifically.	Inherits naive EVs' fragility, plus loading/engineering steps that are themselves yield-loss and stability-risk steps not yet characterized for any human-relevant process.
Naive/unmodified MSC or neural stem cells, delivered intranasally (not as exosomes)	Live cell therapy	More human trials, larger cumulative n, than the EV rows: NEST study (NCT02795052, MSCs, broad neurological indications, IV and intranasal arms, n≈500 planned, recruiting 9+ years with no results published yet); Boika et al. 2020 (Parkinson's disease, Phase II/III, n=12, IV alone vs. IV+intranasal, "significant decrease in severity of motor and nonmotor symptoms" at 1–3 month follow-up); Lv et al. 2023 (NCT03005249, allogeneic neural stem cells, ischemic/hypoxic cerebral palsy in children, Phase I/II, n=25, completed, improved GMFM-88 score, increased grey-matter volume on imaging — the highest-quality-rated trial in a 2026 systematic review of this whole field, Front Aging Neurosci).	No approved intranasal cell product for any neurological indication. Whole-cell products sit under the same HCT/P exclusion logic as EVs when the cells are more-than-minimally-manipulated (culture-expanded), which MSC therapies typically are — full biologic/BLA pathway expected, same as EVs.	A fundamentally different stability problem than EVs: living cells must remain viable, which as a practical matter forecloses lyophilization and most room-temperature presentations outright. Cryopreserved storage and a thaw-and-use (not multi-week multi-dose bottle) administration model is the realistic ceiling — hard to reconcile with a consumer-facing, shelf-stable, multi-dose product format.
MSC secretome / conditioned medium (cell-free)	Secretome	One human intranasal trial found, non-neurological: NCT02192736 (allogeneic umbilical-cord MSC-derived "mesenchymal trophic factors," asthma, Phase I/II, n=20, weekly dosing ×4 weeks, reported as safe and associated with improved pulmonary function and quality of life). For a cognitive/neuro claim specifically, the evidence is preclinical only — intranasal MSC secretome has shown effect in rodent models of perinatal asphyxia, Alzheimer's, and addiction/relapse, but no human neuro trial was found. Separately, Nagoya University's Katagiri group has two published human studies of MSC-conditioned medium (see `japan-asset-longlist.md` — a private working document, referenced here only for the science, not the licensing context) — but by local injection for bone regeneration, not intranasal, and not a cognitive endpoint. Neither is directly transferable precedent.	Same regulatory logic as EVs and cells — a secreted, cell-derived product, not an HCT/P, expected to sit under PHS Act §351. No approved comparator found.	Not a single defined particle population — a complex, biologically variable mixture, which is itself a release-specification problem (see the same issue flagged for a different secretome product in `japan-asset-longlist.md`'s dossier). Likely more freeze-thaw-tolerant than intact EVs since there's no vesicle membrane integrity to preserve, but this is a reasonable inference from mechanism, not a cited stability finding — recorded as open.
Intranasal insulin	Approved small protein, repurposed	The deepest and most controlled human cognitive dataset of any candidate on this list, by a wide margin: SNIFF (NCT01767909), Phase II, ~240 participants, 12 months' treatment, mild cognitive impairment/Alzheimer's, randomized placebo-controlled. Reported outcomes include slowed white-matter-hyperintensity progression and favorable CSF biomarker changes; a 2026 pooled review notes results vary by sex, genotype (APOE), insulin formulation, and delivery device — not a uniform positive result.	Already an approved drug — insulin is a marketed biologic with an existing safety and manufacturing record; the SNIFF-type trials are an off-label repurposing via a nasal delivery device, not a new molecular entity. This is a materially different — and easier — regulatory starting point than any cell- or vesicle-derived candidate on this list.	Small, single, well-characterized protein (~5.8 kDa for human insulin) with decades of industrial manufacturing, formulation, and cold-chain precedent from its approved injectable use. The best-characterized stability profile of any candidate here, though nasal-specific device and permeation-enhancer formulation work (to get a large enough dose across the mucosa) is a separate, real question not resolved by the injectable precedent.
Intranasal oxytocin	Approved small peptide, repurposed	A large body of human trials for social/cognitive endpoints, autism spectrum disorder in particular, with genuinely mixed results — e.g., the NEJM-published pediatric ASD trial (2021) found no significant between-group difference in social/cognitive functioning over 24 weeks, while smaller trials (Breath Powered device, adults with ASD) report dose-dependent positive effects. This is the candidate with the most direct evidence that a nasal peptide can fail to show a cognitive effect in a well-run trial — worth weighing, not just the positive results.	Was FDA-approved (Syntocinon Nasal Spray, Novartis, approved 1960s as a lactation aid) and discontinued in 1997 for commercial, not safety, reasons — still marketed in Europe today. A real approved-product regulatory pathway exists and has been walked before, even though the current product isn't on the US market.	Small cyclic peptide (~1 kDa). Real stability data exists from WHO-prequalified manufacturing: unopened product stable refrigerated (2–8°C), and one study found content "hardly affected" by weeks at 25–30°C — meaningfully more room-temperature-tolerant than any EV or cell candidate on this list.
NAD⁺ precursors (NMN, nicotinamide riboside)	Small molecule	No intranasal human trial found at all. All located human cognitive data is oral: a 12-week NR trial in amnestic MCI (Phase II pilot), and an NR trial in long-COVID that raised NAD⁺ levels but did not improve cognition, fatigue, sleep, or mood versus placebo. Evidence for a cognitive effect specifically is weak even in the oral route this candidate actually has data in.	Marketed as a dietary supplement in the US, not as a drug — a materially different, non-BLA regulatory starting point, but one that would need to change entirely if any disease-treatment claim (like Signal's cognitive-function framing) were made, since a disease claim moves a supplement into drug regulation.	Small, stable, inexpensive molecule by comparison with every other row — but has no nasal-delivery precedent to draw a nasal-specific stability profile from at all.
Reference: approved nasal protein/peptide precedent, non-cognitive (calcitonin-salmon, Miacalcin)	Approved small protein, benchmark only	Not a candidate active for this product — included as the benchmark for what "approved and marketed" actually looks like for a nasal-delivered protein. FDA-approved for postmenopausal osteoporosis; label specifies unopened storage at 2–8°C and in-use room-temperature stability for 30–35 days once opened, in an upright position.	Fully approved (not investigational), decades of post-market history.	32-amino-acid synthetic polypeptide (~3.4 kDa). The clearest existing example of a stable, marketed, room-temperature-in-use nasal protein spray — the practical target every biologic candidate above is implicitly being measured against.

Four-axis scoring

Per the current brief's selection criterion. Mechanism is scored first and weighted as a gate, not one factor among four — a candidate that cannot show a mechanism ranks below one that can. Verdict is traceable to the four scores, not asserted independently of them.

Active	Mechanism	Human-relevant evidence	Manufacturability	Regulatory tractability	Verdict
Naive EVs	Weak/undefined. No demonstrated single-vesicle delivery mechanism in humans; the leading proposed mechanism (miRNA cargo transfer) is itself contested by an unresolved stoichiometry problem — see `mirna-cargo.md`, which finds most preparations carry far less than one copy of a given miRNA per vesicle on average.	Moderate, but thin. Real cognitive-specific precedent (Xie et al., n=9) — the strongest disease-specific human data of any candidate — but open-label, uncontrolled, unblinded.	Weak. None of the seven CDMO-readiness questions is currently answerable (`cdmo-readiness.md`); worst stability profile in this comparison.	Weak. Full BLA pathway, no approved comparator of any kind, excluded from the lighter HCT/P track.	Excluded — no longer the declared development focus (see Development focus, above); weakest of any candidate compared on 3 of 4 axes.
Engineered/loaded EVs	Moderate on intent, unverified in practice. A specific cargo can be named by design, but inherits naive EVs' delivery-mechanism uncertainty on top of it.	None. Zero human intranasal precedent, rodent/in vitro only.	Weakest. Inherits EV fragility plus an unvalidated loading step.	Weakest. No precedent for how FDA treats the combination-product question.	Excluded for now — starts from zero human data on every axis that matters most.
Whole-cell (MSC/NSC) intranasal	Weak/undefined, same order of uncertainty as naive EVs — paracrine/trophic mechanism proposed, not demonstrated at single-cell resolution.	Moderate-strong on volume, larger cumulative n than EVs (NEST, Boika et al., Lv et al.), but split across non-cognitive indications.	Weak, categorically — live-cell viability forecloses lyophilization and most room-temperature presentation, a structural mismatch with a shelf-stable consumer product regardless of how the evidence develops.	Weak. Same full-BLA logic as EVs.	Excluded for this product format — the manufacturability mismatch is structural, not a gap that more data closes.
MSC secretome/conditioned medium	Weak-moderate. A named trophic-factor mechanism class (e.g. IGF-1/VEGF/TGF-β1/HGF-type factors) is more individually characterized than EV cargo, but the product is a variable mixture, not one testable target.	Weak for a cognitive claim — one human intranasal trial exists and it's non-neurological (asthma); cognitive/neuro evidence is preclinical only.	Weak. Not a single defined particle population, which is a release-specification problem the EV rows don't even share in the same form.	Weak. Same secreted-product logic as EVs, no approved comparator.	Declared development focus — mechanism story is marginally better than naive EVs, everything else is comparable or worse; the focus is a business decision, not a score, and the scores are reported as they fall. See Development focus, above.
Intranasal insulin	Strong. Defined CNS insulin-receptor pharmacology.	Strong. SNIFF (NCT01767909), n≈240, RCT, 12 months, the target population — see Comparison table for the documented heterogeneity by sex/genotype/device.	Strong. Approved biologic, established industrial manufacturing.	Strong. Off-label device repurposing of an approved drug.	Benchmark — highest four-axis score of any candidate compared, not adopted. Not a regenerative claim; tracked against the declared focus, see Benchmarks, above.
Intranasal oxytocin	Strong. Defined oxytocin-receptor pharmacology.	Moderate, genuinely mixed. Large trial base, but includes a well-run RCT (NEJM 2021) that found no effect — the clearest documented case of a nasal peptide failing a cognitive/social endpoint in this comparison.	Strong. WHO-prequalified manufacturing precedent, notably room-temperature-tolerant.	Moderate-strong. A regulatory pathway was walked before (Syntocinon, discontinued for commercial not safety reasons) but the product isn't currently on the US market.	Secondary — mechanism and manufacturability are strong, but evidence quality is the weakest-documented of the three non-EV pharmacological candidates.
NAD⁺ precursors	Moderate. NAD⁺ salvage pathway is a defined biochemistry, but with no nasal-specific pharmacokinetics at all.	Weak. No intranasal human trial of any kind; oral cognitive evidence is itself weak or null.	Strong. Small, stable, inexpensive molecule.	Weak where it matters. Currently an unregulated supplement, but a cognitive-function disease claim forces reclassification as a drug — the regulatory advantage doesn't survive contact with the product's own claim.	Excluded — the one axis where it scores best evaporates exactly when the product needs it.

Not assessed this run

Cost and supply availability, and whether candidates can be combined, are still deliberately out of scope. A candidate's molecular fragility interacts directly with what excipient system could stabilize it, so every verdict above should be read as provisional to formulation-design.md.

Excipient and pH constraints are now partly answered, for the exosome candidate only — no longer the declared focus. formulation-design.md was created on 2026-09-03 and works the excipient question for the exosome active, not for MSC secretome (exosome-rich conditioned medium); that work has not been redone since the focus moved, and is flagged as open in that document: measured nasal pH is 6.27–6.40 (Washington et al. 2000) against a neutral-pH EV stability optimum, which is a smaller conflict than expected and is already bridged by an approved product at pH 7.2 (SPRIX). Two findings there bear directly on this comparison and make the EV rows' manufacturability and regulatory scores harder, not easier: every component of the best-evidenced EV stabilisation buffer — trehalose, human serum albumin, HEPES, sucrose, poloxamer 188 — has zero nasal precedent in the FDA Inactive Ingredient Database, and the only permeation enhancer with US nasal precedent (n-dodecyl β-D-maltoside, in TOSYMRA at 0.2%) is a detergent that would be expected to dissolve the active. The equivalent excipient work has not been done for the peptide, secretome, or NAD⁺ rows, so this column is filled for one candidate, not for the comparison. The four-axis scores above are unchanged by it — this run did not rescore anything.

Observations worth flagging now

The four-axis criterion, applied honestly, still scores at least one benchmark candidate above the declared focus (MSC secretome, exosome-rich conditioned medium). Insulin scores higher on mechanism, evidence, manufacturability, and regulatory tractability, but is not a regenerative active and would not support the "nose signals the brain to repair itself" framing the focus candidate is marketed under. This is reported as a comparison finding, not softened toward the focus, per the brief's rule that a declared focus must never change a score. The operator has already made the scope decision; what stays open is whether the marketed mechanism claim is one the focus candidate's own weak mechanism score can actually support — not this agent's call.
Different candidates lead on different axes, which is exactly why scoring separately matters. EVs have the specific cognitive/Alzheimer's precedent (Xie et al.) but the worst mechanism, manufacturability, and regulatory profile of the field. Whole-cell therapies have more aggregate human trials than EVs but a manufacturability mismatch that may be incompatible with any shelf-stable product format at all.
The field has more human intranasal EV trials than the previous spec run had found — three, not one. sgl-001-spec.md was built around Xie et al. as the only human anchor; NCT06598202 and NCT07105371 are new information from this run's search and should be cross-checked into that spec the next time it's updated, including the fact that NCT06598202 is the only controlled human intranasal EV trial found anywhere in this comparison.
Engineered/loaded EVs have no human intranasal precedent at all. If a future run considers this direction, it should be scored as starting from zero human data, not as an incremental step up from naive EVs.
The Nagoya secretome work referenced above lives in a private working document (japan-asset-longlist.md), not because the science is confidential, but because that document's licensing/BD context is. Citing the published science here, in a public artifact, without the BD framing, is intentional and within the rules governing that file.

Sources consulted

Xie X, et al. Gen Psychiatr. 2023;36(5):e101143 (NCT04388982) — already the primary human EV anchor in sgl-001-spec.md; re-cited here for the active comparison specifically.
NCT06598202 — umbilical-cord MSC exosomes, ALS, randomized double-blind placebo-controlled, via the systematic review below (not independently re-fetched from the registry this run).
Prodromos, 2025 (NCT07105371) — allogeneic UC-MSC exosomes, motor-neuron and movement disorders, via the systematic review below.
Intranasal administration of stem cells and their derivatives for neurological and respiratory disorders: a systematic review of human clinical trials, Frontiers in Aging Neuroscience, 2026 — the source for the EV, cell, and secretome trial data above (NEST/NCT02795052, Boika et al. 2020, Lv et al. 2023/NCT03005249, NCT02192736), fetched and summarized this run, not independently verified against each underlying primary registry entry. Flagging per this agent's own standard: a systematic review is a strong secondary source, not a substitute for the primary registry/publication if any of these trials becomes a serious candidate.
Georgetown University Medical Center press materials and NCT01767909 (SNIFF) — intranasal insulin, Alzheimer's/MCI.
Ahmadian S, et al. Stem Cell Res Ther. 2024;15:453 — freeze-thaw particle loss data, already cited in sgl-001-spec.md; re-cited for the EV stability row.
NEJM 2021 (Sikich et al.), and the Breath-Powered-device autism oxytocin trial (Transl Psychiatry 2017) — intranasal oxytocin cognitive/social outcomes.
Syntocinon Nasal Spray prescribing information / regulatory history; oxytocin WHO-prequalified stability data (PMC7646793).
Nicotinamide riboside MCI pilot trial and long-COVID RCT (eClinicalMedicine, 2025) — oral NAD⁺ precursor cognitive data; no intranasal human trial was found for any NAD⁺ precursor.
Miacalcin (calcitonin-salmon) nasal spray label, FDA — approved-nasal-protein reference row.

This run's sourcing for the four-axis scoring, disclosed plainly rather than left implicit. The mechanism, manufacturability, and regulatory-tractability scores for the EV, whole-cell, and secretome rows are derived directly from already-cited artifacts in this repo — cdmo-readiness.md, mirna-cargo.md, jurisdiction-map.md, sgl-001-spec.md — not re-fetched this run. The same scores for insulin, oxytocin, and NAD⁺ precursors draw on established receptor pharmacology and each product's public approval history (insulin- and oxytocin-receptor signalling, the Syntocinon and NAD⁺-supplement regulatory histories) rather than a fresh primary-source search specific to this run — these are treated as settled pharmacological and regulatory fact, not as this agent's own prior inference, but the distinction is recorded here so a future challenge pass knows exactly what was and wasn't independently checked this run.

Run history

2026-09-04Brief and artifact now agree. Set exact operator-specified notes on the focus and excluded entries in formulation.json, and folded the same purification-cost rationale into active-candidates.md's Development focus prose so the source document carries the reasoning, not only the derived JSON. No score, class, or entry name changed.
2026-09-04Every numeric axis score and every fact in the Comparison and Four-axis tables is unchanged — only the focus/excluded labels and the surrounding narrative moved. Secretome scores weak-to-weakest on 3 of 4 axes as the new focus; both benchmark candidates (insulin, oxytocin) now score above it on all four axes, not just three.
2026-09-01Restructured 'Current lead' into 'Development focus' plus 'Benchmarks'; the scoring table's axis cells are unchanged. EVs still score weakest on 3 of 4 axes and insulin still scores highest on all 4 — the declared focus changed how the artifact frames those facts, not the facts.
2026-09-01Current lead is intranasal insulin, not extracellular vesicles — strongest on all four axes but not a regenerative active, a tension flagged for the operator rather than resolved.
2026-08-28First formulation pass; excipient system, device, and a final verdict deferred to a later run.