Dyno Therapeutics

Phase A — Understand the business

Lens 1 · Company Overview

Dyno is an AI-driven AAV-capsid-engineering company — it designs the delivery vehicle for gene therapy, not the gene therapy itself. The core insight of the whole company: in gene therapy the genetic payload is increasingly the easy part; the hard, unsolved bottleneck is delivery — getting an adeno-associated virus (AAV) to the right tissue, at a low enough dose, while evading pre-existing immunity. a16z framed the problem precisely when it led the Series A: AAVs suffer "insufficient payload delivery, manufacturing difficulties, safety concerns at high doses, and pre-existing patient immunity," and traditional optimization fails because "one false step can result in a precipitous fall that breaks the capsid" and improving one property degrades another. Despite 33 FDA-approved cell/gene therapies, only ~6 deliver genes in vivo — the rest are ex-vivo — which is the size of the delivery problem.

The platform — CapsidMap → LEAP → "Super-Darwinian evolution." Dyno's engine is a closed loop: (1) DNA-barcoded multiplexed libraries create billions of AAV-capsid variants at once; (2) high-throughput in-vivo screening + next-gen sequencing measures each variant's actual tissue-targeting / immune-evasion / manufacturability "fitness"; (3) machine-learning models trained on that proprietary data predict which untested sequences will win and propose the next library. The seminal data asset is a 2019 Science paper that mapped "every single codon substitution, insertion, and deletion of AAV2" into a complete fitness landscape; a 2021 Nature Biotechnology collaboration with Google Research showed "richer machine-learning models could generate better AAV designs". The platform has since been re-branded LEAP ("Low-shot Efficient Accelerated Performance").

The business model — "Intel inside," not a drug company. This is the single most important strategic fact. Dyno deliberately does not run its own clinical trials. It licenses engineered capsids to pharma partners, who take on all preclinical/clinical/commercial work, and Dyno collects upfront + research funding + option/license fees + development & sales milestones + royalties. The thesis: Kelsic estimates "less than 100 engineered capsids" can solve most delivery challenges, so being the early IP owner of the best ~100 capsids → "hundreds or even thousands of AAV gene therapies entering the clinic, each using a Dyno capsid" → a royalty annuity broader than any single owned blockbuster. It is the ARM/Intel-of-gene-therapy bet, not the Sarepta bet.

Customers (named pharma licensees): Novartis (ocular), Sarepta (muscle), Roche (CNS/liver, then neurological), Spark Therapeutics (a Roche company, liver), Astellas (skeletal & cardiac muscle), plus tech partner NVIDIA. Suppliers / inputs: GPU compute (NVIDIA/BioNeMo), DNA synthesis + NGS, non-human-primate testing capacity, and — newly — a CDMO manufacturing partner, Trisk Bio (Nov 2025). Competitors: Voyager Therapeutics (TRACER directed-evolution platform), 4D Molecular Therapeutics, Capsida Biotherapeutics, plus pharma in-house capsid groups.

Contract terms: classic platform-licensing economics (upfront → research funding → option/license fee on capsid selection → milestones → royalty). Exact royalty rates and seat-level terms are undisclosed — n/a.

Lens 2 · Supply Chain

For a capsid-design company the "supply chain" is data-generation inputs → Dyno's design loop → the licensed capsid → the partner's gene therapy → the patient. Named stakeholders along the chain:

Upstream inputs:

• Compute / AI infra — NVIDIA. A formal technology partnership (May 2024) to train foundation-scale models on NVIDIA BioNeMo; extended in 2026 into the Dyno Psi-Phi agentic protein-design suite (built on BioNeMo NIMs incl. OpenFold3). Both a compute dependency and a credibility signal.
• Wet-lab data factory — Dyno's own high-throughput in-vivo screening. The differentiated input is proprietary functional fitness data on billions of variants — the thing open structure-prediction models (AlphaFold/ESM) do not give you. This is the moat input (Lens 3).
• DNA synthesis + NGS — barcoded library synthesis and sequencing read-out; commodity-ish but throughput-critical.
• Non-human-primate (NHP) testing — the gold-standard preclinical proof; Dyno reports NHP data on every flagship capsid (e.g. Dyno-9zh transduces "up to 50% of neurons in premotor cortex at 3×10¹³ vg/kg in NHPs," Dyno-n96 "~74% of skeletal myofibers and ~16% of cardiomyocytes in NHPs").

Dyno (the company): the LEAP design loop + the proprietary fitness datasets + the engineered-capsid IP estate + a B2B licensing GTM org.

Downstream (named): the licensed capsid → Roche / Spark (CNS, liver, neurological), Astellas (skeletal & cardiac muscle), Novartis (ocular, deal completed 2024), Sarepta (muscle, deal completed 2025) → each partner's gene-therapy drug program → the patient. Manufacturing: the capsid then has to be produced at GMP scale — historically the field's worst bottleneck — which is why the Trisk Bio (Nov 2025) "strategic manufacture" partnership matters. New "Dyno Frontiers" model: Dyno is also pairing its capsids with external payload developers — e.g. Epicure (an Allen Institute spinout doing promoter/cell-type-targeting design) — to assemble more complete delivery+payload packages.

Chokepoints / single-source dependencies: (1) NHP capacity — the rate-limiter on validating any new capsid. (2) GMP manufacturing — a sector-wide chokepoint Dyno doesn't own (mitigated by Trisk). (3) The partner channel itself — Dyno has no direct route to a patient; if pharma exits AAV (and several just did — Lens 12/13), Dyno's entire demand side contracts. (4) Talent — a few dozen rare ML-×-virology engineers out of the George Church lineage. Names along the chain — done: NVIDIA, Trisk Bio, Roche, Spark, Astellas, Novartis, Sarepta, Epicure.

Lens 3 · Competitive Advantages (moats)

Verdict: a genuine, compounding data-and-IP moat — the strongest in its niche — but pointed at a shrinking market and threatened by its own customers' ability to insource.

What is genuinely defensible:

• The proprietary fitness-data flywheel. Open models (AlphaFold3, ESM) predict structure; they do not tell you which capsid actually targets muscle at low dose and dodges neutralizing antibodies in vivo. Dyno's billions-of-variants barcoded screening data is exactly that labeled functional data. Per Century of Bio, "each new capsid design project should be cheaper and should produce data that accelerates the next project" — a real economies-of-data loop. This is the deepest leg.
• Demonstrated performance lead over human design. At ASGCT 2025 Kelsic showed "capsids designed with LEAP dramatically outperformed capsids designed by humans without AI assistance". Concrete fold-improvements: bCap1 ~100× over AAV9 (brain), eCap1 ~80× over AAV2 (eye); Dyno-yp2 ~29× lower liver vs AAV9 and ~80× lower vs a benchmark TfR capsid; Dyno-n96 ~25× lower dose vs existing DMD therapies. These are the moat made measurable — lower dose is directly the answer to AAV's dose-driven toxicity (Lens 12).
• The IP land-grab. If "<100 capsids" really do cover most of the delivery map, the early owner of the best ones has a patent moat that compounds as each enters a partner's clinic.
• Pedigree as a moat-adjacent asset. Born from George Church's Harvard lab (the dominant synthetic-biology lineage); Kelsic personally "measured the first comprehensive fitness landscape of the AAV capsid and co-discovered the AAV MAAP gene". Hard to assemble that team.

Bargaining power — who needs whom more: mixed, and improving for Dyno on the capability axis but weak on the channel axis. Pharma "[hasn't] been able to crack [capsid design] on their own so far" — which is why five majors licensed rather than built. But each licensee is a multi-hundred-billion-dollar company that can fund an in-house capsid group, multi-source (Voyager/4D/Capsida), or simply exit AAV — and in 2025 Biogen, Pfizer, Vertex and Takeda did exactly the last thing. Dyno's leverage is its data lead + IP; its weakness is total dependence on a partner channel it doesn't control.

The structural ceiling (same as every picks-and-shovels biotech): by not owning the drug, Dyno captures milestones + royalties, not the full value of the therapy. Century of Bio flags the precedent bluntly: antibody-licensing companies are "an order of magnitude smaller" than vertically integrated drug owners. That is the bull/bear crux (Lens 12).

Lens 4 · Segments

No segment-level financials are disclosed — Dyno is private and reports nothing. segments.csv is empty; every figure here is qualitative ``. There is no revenue/EBITDA split to source — n/a for all segment numbers. What can be characterized is the capsid portfolio by target tissue (the operational "segments"):

• Trend & cause: the center of gravity has shifted to CNS — Roche's 2024 neurological expansion ($50M upfront, the largest disclosed single upfront) and the JPM-2026 brain-delivery showcase make CNS the growth segment; muscle is validated-but-commoditizing (and now under the Sarepta safety cloud); ocular/Novartis is completed (a finished engagement, not recurring). Direction: accelerating on capsid count and tissue breadth (eye → muscle → CNS → liver) and on the frontier (a 2026 push beyond AAV into general protein-binder design, Dyno Psi-Phi — Lens 5/8); the financial slope is unobservable.

---

Phase B — Measure performance

+private overlay: Phase B swaps the SEC-grounded earnings lenses for funding/valuation trajectory, traction signals, and cap-table quality. All ``, unaudited.

Lens 5 → Funding & valuation trajectory (replaces Earnings Result)

The priced-round history, seed → latest:

• Total equity raised: ~$109M across the two priced rounds.
• No priced round since the May-2021 Series A — i.e. ~5 years without a new equity mark. No Series B is disclosed. This is the defining financial fact and it is double-edged (Lens 11/13).
• The actual funding story is non-dilutive partner cash. Disclosed upfronts/options/licenses: Roche $50M upfront (Oct-2024) + CNS license exercise (Jan-2025, fee undisclosed); Astellas $18M upfront (Dec-2021) + $15M skeletal-muscle license exercise (2026); Sarepta "$40M in upfront, option and license payments during the research phase"; Novartis undisclosed upfront + research funding + license fees. Aggregate potential biobucks across deals: a16z cited "over $4 billion in aggregate" (Sarepta + Novartis + Roche/Spark); Astellas adds "up to $1.6B"; Roche's 2024 expansion adds ">$1B". Realized cash is a fraction of that headline — but the point is that a licensing platform can fund itself through a capital winter on milestones, which an owned-pipeline biotech cannot.
• Burn / runway: n/a — not disclosed. With ~$109M equity raised back in 2021, ~50 employees as of 2021 (since grown — exact 2026 headcount n/a), and five years elapsed, the company is almost certainly running materially on partner milestone/license revenue rather than residual Series A cash — a structurally healthier position than a cash-burning clinical biotech, but unverifiable without a fresh raise or disclosure.

Lens 6 → Founder & ecosystem signal (replaces Earnings Calls sentiment)

No earnings calls exist. The analog — founder communication + ecosystem tone, and how it has shifted:

• Eric Kelsic is a prolific, technically-credible public communicator (Nucleate Insights, Benchling "Transcribed," Big Think, BIO 2026, Synthetic Biology Summit 2026, JPM 2026) and stays relentlessly on one message: delivery is the bottleneck, AI-designed capsids are the unlock, and Dyno is the horizontal platform that arms everyone.
• Tone shift 2024→2026 — from "platform validation" to "delivery agility" + safety realism. At JPM 2026 the framing was notably more sober and mechanism-plural: "Very soon, there's going to be a safe demonstration of CNS delivery… the only question is really which mechanism is going to be first," and "not one mechanism is the best for everything". The emphasis on multiple BBB-crossing mechanisms rather than one universal vector is a direct, healthy response to the Sept-2025 sector death (a rival's TfR vector — Lens 10/13). This is maturation, not spin.
• A 2026 platform broadening signal: the launch of Dyno Psi-Phi / Psi-1 (open-weight protein-binder design, beyond AAV) at NVIDIA GTC 2026 — even shipping Claude Code Skills for authenticated API access — reads as Dyno positioning its ML stack as a general protein-design capability, not only capsids. Bullishly: optionality + a developer-ecosystem land-grab. Bearishly: possible diversification away from a stalling core AAV market (Lens 13).

Lens 7 → Cap table & peer / mechanism comps (replaces Comps)

Cap-table quality: a tier-1 syndicate — a16z (Series A lead; its bio/health franchise), GV (Alphabet), Lux Capital, Casdin Capital (a specialist life-sciences crossover — the closest thing to an IPO-track investor on the table), Obvious Ventures, plus founding seed leads Polaris and CRV and KdT Ventures. No disclosed Fidelity / T. Rowe / Coatee-style public-crossover mark and no Series B — so by the private-overlay heuristic, Dyno is not currently IPO-proximate; Casdin's presence is the only crossover-adjacent tell, and it is from 2021. Secondary marks / current valuation: n/a — not disclosed.

Peer comparison — by mechanism/approach, not by P/E (multiples n/a for all):

The comp set frames the strategic fork: Voyager and 4D are public and own pipelines (capturing drug upside); Dyno is private and pure-licensing (capturing royalties only). Dyno's capital efficiency (5 majors on ~$109M equity) is the flip side of its capped ceiling. No revenue multiple is sourceable for any of these as a capsid-platform read — n/a; do not fabricate one. (Voyager/4D public EV/Sales exist but are dominated by their owned-pipeline economics, not the capsid-licensing line — not a clean comp for Dyno.)

Lens 8 → Funding & product / data catalysts (replaces Stock-Price Catalysts)

No stock, so "what moves the name" = funding, partnership, and data events:

• Nov 2018 — $9M seed (Polaris/CRV) → company exists, Church-lab-credible.
• May 2020 — emerges from stealth with Novartis (ocular) + Sarepta (muscle) deals simultaneously → instant commercial validation.
• Feb 2021 — Nature Biotechnology paper (with Google Research) → scientific credibility.
• May 2021 — $100M Series A (a16z, ~$500M post) → the clearest value step-up; capital to scale platform + add partners.
• Dec 2021 — Astellas muscle deal ($18M up, up to $1.6B).
• 2024 — Roche neurological expansion ($50M upfront, >$1B); NVIDIA/BioNeMo technology partnership → the largest single upfront + an AI-compute halo.
• Jan 2025 — Roche exercises the CNS capsid license — the first commercial validation of an AI-designed AAV capsid for neurological gene therapy. The most important de-risking event to date.
• Sept 2025 — negative read-through: a child dies in Capsida's BBB-crossing (TfR) AAV trial → casts a safety cloud over the exact CNS mechanism Dyno also pursues (yp2 is TfR).
• Nov 2025 — Trisk Bio strategic-manufacturing partnership (de-risks the GMP chokepoint).
• 2026 — Astellas exercises the skeletal-muscle capsid license ($15M) → second commercial license exercised; Dyno Psi-Phi / Psi-1 launched at NVIDIA GTC (protein-binder expansion); JPM 2026 brain-delivery data (Dyno-yp2); ASGCT 2026 new capsids (Dyno-9zh CNS, Dyno-n96 muscle).
• Pattern: Dyno's "stock" reacts to (a) marquee partner deals/expansions, (b) license-exercise events (capsid actually selected for a program — the real value step), and (c) head-to-head NHP performance data. The next true catalysts: a partner's Dyno-capsid program entering the clinic with clean safety (the thesis-maker), and/or a Series B / fresh priced mark (the funding tell). The big tail risk: a Dyno-capsid clinical safety event, given the sector backdrop.

---

Phase C — Judge people & books

Lens 9 · Management

Eric D. Kelsic, PhD (Co-founder & CEO). The defining asset. Physics at Caltech, then trained under George Church at Harvard, where he "led the team that developed the technology underlying Dyno's AI-powered capsid engineering platform," "measured the first comprehensive fitness landscape of the AAV capsid," and "co-discovered the AAV MAAP gene". This is a rare founder-scientist who literally invented the core IP, not a hired operator. Skin in the game: founding CEO since Nov 2018 (~7.5 years), presumably a large equity holder (exact % n/a — not disclosed).

Co-founders: George Church (the synthetic-biology figurehead — pedigree + network), Adrian Veres, Sam Sinai (ML/evolutionary modeling — "Super-Darwinian evolution"), Alan Crane (Polaris Partners entrepreneur-partner — the seasoned biotech-builder), and Tomas Bjorklund (Lund University — AAV/CNS expertise). The founding logic — Church-lab AAV science + ML + a Polaris operator — is precisely the combination this category rewards.

Capital-allocation history: too early / private to judge on ROE/ROIC. What is visible is disciplined capital efficiency and a non-dilutive funding instinct — five pharma majors signed on ~$109M of equity, and the company has funded ~5 additional years substantially on partner milestone/license cash rather than serial dilution. The deliberate no-owned-pipeline choice is itself a capital-allocation philosophy: avoid concentrating clinical risk, monetize the platform horizontally.

Red flags (management): none material found in public sources — no related-party scandals, no promotional over-claiming beyond normal biotech press, no governance smells. The only soft flag is strategic, not ethical: the licensing-only model may cap value (the bear's core charge), and the 2026 broadening into general protein binders could read as either smart optionality or a hedge against a stalling AAV core (Lens 13).

Archetype: founder-scientist, mission-driven, technically dominant — the right archetype for a frontier platform at this stage, with a credible commercial operator (Crane/Polaris) in the founding DNA. Kelsic is a builder-explainer, not a hype-merchant — his JPM-2026 safety realism on CNS mechanisms is evidence of that.

Lens 10 · Forensic Red Flags

Dyno is private and unaudited — there is no income statement, balance sheet, or cash-flow statement to forensically examine. The classic checklist (revenue recognition, receivables vs revenue, SBC, goodwill, leases, related-party games) is n/a — no audited financials disclosed. The honest forensic posture: you are trusting management's selective disclosures (fold-improvement claims, deal headlines, NHP percentages) with no auditor and no GAAP. Specifically un-verifiable and worth flagging:

• The fold-improvement and biobucks numbers are self-reported / best-case. "~100× / ~80× / ~29× / ~25×" are company press-release figures from NHP or mouse studies — directional, not independently audited, and preclinical (mouse/NHP ≠ human). The "$4B+ aggregate / $1.6B / $1B+" deal values are biobucks — almost all contingent milestones that mostly will never be paid; realized cash is a small fraction.
• License exercises are the only hard value events — Roche (Jan-2025) and Astellas ($15M, 2026) are the figures that are real money for a selected capsid; treat everything else as option value.
• Revenue / margin / burn / runway / current valuation: all n/a.

Regulatory findings (required sub-section).

• SEC (EDGAR EFTS — LR + AAER): none possible. Per regulatory/regulatory-findings.md (generated 2026-06-29), Dyno has no CIK — private, not required to file with the SEC; total SEC findings = 0.
• Non-SEC enforcement (FTC / DOJ / FDA / CFPB / EU) — web search run ("Dyno Therapeutics" (FTC OR DOJ OR FDA OR... ) enforcement): no material enforcement hits against Dyno itself. As a capsid licensor that does not itself sponsor a clinical trial or market a drug, Dyno's direct FDA/clinical-liability surface is limited — the regulatory risk sits primarily with its partners' INDs/trials.
• Sector / partner regulatory context (material, indirect): in 2025 the FDA halted Sarepta's Elevidys after three patient deaths from AAV-associated acute liver failure (AAVrh74 serotype) and revoked the AAV "platform-therapy" designation that had been speeding AAV reviews. Sarepta is a former Dyno partner (deal completed 2025); and Dyno's Dyno-n96 is explicitly an AAVrh74-based muscle capsid — the same serotype family now carrying a boxed warning. Separately, Sept-2025: a child died in Capsida's BBB-crossing (TfR) AAV trial — a rival's vector, not Dyno's, but the same TfR mechanism Dyno's Dyno-yp2 uses. Neither death is a Dyno product — but both land squarely on Dyno's two lead tissue segments (muscle rh74; CNS TfR). This is the most important risk fact in the dossier and it must not be misread as a Dyno-specific enforcement action.
• Item 3 Legal Proceedings: no 10-K exists (private) — n/a.
• Conclusion: No material regulatory or legal findings against Dyno Therapeutics itself — verified via SEC EDGAR EFTS (LR + AAER, zero records, no CIK) and web search across FTC/DOJ/FDA/CFPB + litigation/controversy, as of 2026-06-29. Unaudited per public sources. The real risk is not Dyno enforcement — it is (a) the absence of any audited disclosure, and (b) the AAV sector's deteriorating safety/regulatory backdrop bearing directly on Dyno's lead mechanisms.

---

Phase D — Project & stress-test

Lens 11 → IPO-readiness & path-to-tradeable + capsid rNPV (replaces Forward Projection)

There is no research/private-watch.json entry for dyno-therapeutics to anchor stage/ipo_readiness/catalyst — so I assess from first principles.

IPO-readiness & path-to-tradeable:

• Stage: growth, but unusually aged at stage — Series A done May-2021 (~$109M total), no priced round in ~5 years, funding itself on partner milestones. That is atypical: either disciplined (doesn't need dilutive capital) or stalled (couldn't raise a competitive Series B into the AAV winter) — and from outside you cannot tell which. ``
• IPO-readiness: ~2 / 5 (1=seed … 5=S-1-imminent). Rationale: a platform-licensing company can go public on royalty/milestone economics, but Dyno has no fresh crossover round, no disclosed revenue, no owned clinical asset to headline an S-1, and is staring at the worst public-market window for anything AAV-adjacent in a decade (Biogen/Pfizer/Vertex/Takeda exits; Sarepta boxed warning). Not IPO-proximate in 2026. ``
• Milestones that would unlock a tradeable event (the be-early watch-list): (1) a partner's Dyno-capsid gene therapy entering the clinic with clean human safety — the single fact that would re-rate the whole AAV-delivery thesis; (2) a fresh priced round (Series B) — especially with a true public-crossover (the funding/valuation tell that's been absent 5 years); (3) escalating royalty-bearing license exercises (Roche + Astellas are two — a third/fourth converts the "annuity" story from claim to evidence); (4) Dyno Psi-Phi traction as a general protein-design business (a second leg that doesn't depend on AAV).
• More-likely outcome than IPO: strategic acquisition. Natural acquirers: a gene-therapy-committed pharma wanting to own the delivery moat outright (Roche/Astellas are already deep in — the relationship cuts both ways), a tools/AI-bio platform (NVIDIA-adjacent, or a Schrödinger/Recursion-type), or a CDMO/tools major (Danaher, Thermo). For a pure-platform delivery IP estate with proven partner pull, M&A is the base-rate exit.
• Investability TODAY: not directly investable — private, no secondary access surfaced, no priced mark since 2021. This is a watch / private-frontier name — banked for a fresh round, a clean-safety partner clinical entry, or an S-1.

Capsid "rNPV" intuition (the +clinical lens, qualitative — every input ``): Dyno's economic value ≈ Σ over licensed capsids of (partner program peak sales × royalty rate × probability the program reaches market). Each input is dark — royalty rates undisclosed, partner peak-sales unknowable, and AAV PoS just fell sector-wide on the 2025 safety events. The two exercised licenses (Roche CNS, Astellas muscle) are the only nodes with non-trivial probability mass today; everything else is option value on a de-rating sector. A point rNPV would be false precision — n/a on a number; the directional read is that the option set is broad but each option's PoS just got marked down.

(No forecast.ts create logged — per skill rules the --watchlist/unattended loop skips the forecast-create step, and a private, no-owned-pipeline licensor has no Dyno-owned EPS line or Dyno-owned binary clinical readout to score. The scoreable binary that would matter — "a Dyno-capsid partner program posts clean Phase 1 CNS safety by YYYY" — belongs to the partner, not Dyno; correct to skip the create here and flag it for Connor.)

Recommended private-watch.json seed (for Connor to add outside this loop — NOT written by this run per wave boundaries):

"dyno-therapeutics": { "beat": "ai-bio", "stage": "growth-aged", "ipo_readiness": 2,
 "lead_investors": "a16z, GV, Lux, Casdin, Polaris, CRV",
 "catalyst": "Partner Dyno-capsid clinical entry with clean safety; first fresh priced round / crossover in 5yrs; more royalty-bearing license exercises (Roche+Astellas=2); Dyno Psi-Phi protein-binder traction; possible strategic acquisition (Roche/Astellas/NVIDIA-adjacent/Danaher/Thermo)",
 "risk": "AAV sector safety winter (Sarepta rh74 deaths + FDA platform-designation revoked; Capsida TfR CNS death) lands on Dyno's two lead mechanisms",
 "dossier": "companies/dyno-therapeutics/deep-dive-2026-06-29.md" }

Lens 12 · Bull vs Bear

Bull case. Dyno is building the delivery layer for the entire gene-therapy industry — the "Intel inside" of AAV — and it owns the one thing pharma demonstrably can't build alone: a compounding, proprietary, in-vivo fitness-data flywheel that produces capsids measurably 25–100× better than human or wild-type designs on the metric that matters most — dose. Lower dose is the direct antidote to AAV's central problem (dose-driven liver and immune toxicity — the very thing that just killed patients on old serotypes), so a safety crisis in legacy AAV is, on this reading, a demand driver for better-engineered capsids, not a death knell. Five pharma majors (Novartis, Sarepta, Roche, Spark, Astellas) validated the platform, two have now exercised commercial licenses (Roche CNS Jan-2025, Astellas muscle 2026 — the first AI-designed AAV capsids selected for real programs), and the model is gloriously capital-efficient and self-funding through milestones — which is exactly why it survived a capital winter that bankrupted owned-pipeline peers without a single down round. Add a credible second leg (Psi-Phi general protein-binder design with NVIDIA) and a manufacturing partner (Trisk), and the bull sees the royalty-annuity-on-the-whole-industry compounding for a decade, with a clean strategic-acquisition floor under it.

Bear case (2–3 permanent-impairment risks).

1. The market for the product is actively contracting. Dyno sells picks-and-shovels into AAV gene therapy — and in 2025 Biogen, Pfizer, Vertex and Takeda all exited or wound down AAV programs, Pfizer killed an approved product (Beqvez) for non-uptake, and the FDA revoked the AAV platform-therapy designation after Sarepta's deaths. If the customer base for AAV shrinks, the best capsid in the world has fewer programs to ride. A picks-and-shovels business is only as good as the gold rush.
2. The lead mechanisms just got a safety cloud — and Dyno captures none of the downside protection of owning the molecule. Dyno's two flagship segments are muscle (AAVrh74 — now boxed-warning serotype family, and Dyno-n96 is rh74-based) and CNS (TfR — the mechanism in Capsida's fatal Sept-2025 trial). Dyno didn't cause either death, but a partner's Dyno-capsid program suffering a clinical safety event would impair Dyno's credibility with no owned-asset value to cushion it — it has no drug, only a licensing fee.
3. No fresh priced mark in ~5 years + total financial opacity. No Series B, no disclosed revenue/burn/runway/current valuation. The "self-funding on milestones" story is plausible but unverified; the alternative read — that it couldn't raise a competitive up-round into the AAV winter — is equally consistent with the facts. A picks-and-shovels royalty company with undisclosed economics is a faith-based valuation.

Pre-mortem (18 months out, thesis broke): A partner's Dyno-capsid CNS program hit the same TfR-class safety signal that felled Capsida; pharma's AAV retreat deepened (another major exited); the milestone cash thinned as programs were paused; Dyno was forced into a flat/down Series B (its first mark in 6 years) at terms that signaled the AAV-delivery dream had cooled; and because it owned no molecule, there was no clinical win anywhere in the portfolio to cushion the narrative. It survives — as a strong IP estate acquired by Roche or a tools major at a capability price, not the category-defining platform the 2021 $500M mark implied.

Are multiples too high? Unobservable (private, undisclosed). But the category (AAV) just de-rated hard across the public comps (Sarepta, 4D, Voyager all under pressure), so a 2021-vintage $500M private mark carries real down-round risk into any 2026–27 raise if sentiment doesn't recover.

Contrarian view (what the market refuses to see): the crowd reads the 2025 AAV bloodbath as a sector death sentence and is throwing Dyno out with it. The contrarian read: the bloodbath is a delivery problem, and delivery is the one thing Dyno exists to fix. Every death from dose-driven toxicity on a wild-type serotype is an argument for a 25–100×-more-efficient engineered capsid that works at a fraction of the dose. If AAV survives at all, it survives on better capsids — and Dyno owns the best data for making them. The market is pricing Dyno as "AAV, and AAV is dying"; the contrarian prices it as "the only credible path to AAV not dying." The risk to the contrarian: the market may be right that AAV itself — not just bad capsids — is the dead end (LNP/non-viral and in-vivo editing eat the use cases), in which case Dyno is the best blacksmith in a world going electric.

Lens 13 · Devil's Advocate (short-seller)

Dismantling the bull case:

• You are the arms dealer in a war the generals are leaving. Strip the AI romance and Dyno = (a great capsid-design dataset) + (an IP estate) sold to a customer base that is shrinking in real time. Biogen out, Pfizer out (and killed an approved AAV drug), Vertex out, Takeda out, Sarepta boxed-warned and FDA-restricted. The most dangerous competitor bulls underestimate isn't Voyager or Capsida — it's the substitution of the whole modality: LNP/non-viral delivery and in-vivo CRISPR editing are eating exactly the indications AAV hoped to own. Dyno is optimizing the buggy whip.
• Revenue concentration is total and the tail is contingent. Demand is ~5 pharma logos, two of which (Novartis, Sarepta) are completed/wound-down. And because Dyno owns no molecule, "$4B+ aggregate biobucks" is almost entirely never-to-be-paid milestone vapor — realized value is the handful of upfronts + two exercised licenses. It's a subscription-of-options that the customers can cancel by exiting AAV — which they are doing.
• The moat may be narrower than the fold-numbers suggest. "25–100×" are preclinical mouse/NHP press figures, not human efficacy or safety — and the field has a long graveyard of capsids that dazzled in NHP and disappointed (or killed) in humans. Worse, Dyno's two lead mechanisms are the two with fresh human safety signals (rh74 muscle; TfR CNS). Meanwhile a Roche or Novartis with effectively unlimited budget can stand up its own capsid-screening loop, and Voyager/4D are doing directed evolution that also works.
• Worst capital-allocation / incentive critique: none egregious — but 5 years without a priced round is a yellow flag dressed as discipline. "Self-funding on milestones" can also read "can't get a competitive up-round in this tape." And the 2026 pivot-broadening into general protein binders (Psi-Phi) — open-weighting models, shipping Claude Code skills — looks suspiciously like a new TAM to chase because the core AAV TAM stalled. Optionality, or distraction from a stalling core?
• What must hold for the (2021, $500M) mark: that AAV survives as a large, durable modality and that capsid quality (not modality substitution or manufacturing or reimbursement) is the binding constraint and that pharma keeps licensing rather than insourcing and that Dyno's engineered capsids prove safe in humans, not just potent in NHP. That's four conjunctive bets, and 2025 weakened at least two. If AAV program count disappoints 20–30%, the next round is a down round, the IPO path stays shut, and the outcome compresses to a capability acquisition.
• The single permanent-impairment scenario (and plausibility): a partner's Dyno-capsid program posts a serious human safety event (entirely plausible given rh74/TfR exposure) while the broader AAV retreat continues → Dyno's core value proposition ("our capsids are better and safer") is falsified in the clinic with no owned asset to fall back on, and it becomes a discounted IP sale. Plausibility: moderate — the lower-dose engineering is a genuine safety argument for Dyno and the platform breadth is a real buffer, but the direction of travel (modality substitution + sector exits + safety signals on Dyno's exact mechanisms) is unambiguously against the most expansive 2021 valuation.

Lens 14 · Management Questions (ordered by information value)

1. Has any partner's Dyno-designed capsid entered (or when will it enter) human trials — and what is the dose and the early safety read? (The single fact that converts "potent in NHP" into "the thesis works" — and the whole bet rides on clean human safety, especially after rh74 and TfR signals.)
2. Given Sarepta's rh74 deaths and the FDA pulling the AAV platform designation, why is Dyno-n96 still rh74-based, and what is your evidence that a lower-dose engineered rh74 capsid materially changes the liver-toxicity risk in humans?
3. The Sept-2025 Capsida death was a TfR BBB-crossing vector — your Dyno-yp2 is also TfR. What is your read on whether that is a mechanism-class risk, and how does your "multiple BBB mechanisms" strategy de-risk it?
4. You haven't raised a priced round since the 2021 Series A. Are you default-alive on partner milestone/license revenue — and at what point, if ever, do you raise a Series B, and would you expect an up-round? (The financial tell that's been dark for 5 years.)
5. With Biogen, Pfizer, Vertex and Takeda exiting AAV, how is your new-partner pipeline trending — are you signing new programs or defending an existing book? What does demand actually look like in 2026?
6. What are your royalty rates and how much realized (not biobucks) cash have the deals generated to date — i.e., what fraction of the "$4B+ aggregate" is real?
7. Roche (Jan-2025) and Astellas (2026) exercised licenses. How many additional license exercises are realistically in flight, and on what timeline? (Exercises, not signings, are the value events.)
8. What is your defense when a Roche or Novartis decides to build capsid-screening in-house — what specifically can't they replicate, and for how long?
9. Is AAV itself the right long-horizon bet, or do LNP/non-viral delivery and in-vivo editing eventually substitute for the indications you target? How do you hedge modality risk?
10. Dyno Psi-Phi / protein binders — is this a genuine second business or a hedge against a stalling AAV core? What's the revenue model and is it dilutive to focus?
11. The fold-improvement figures (25–100×) are NHP/mouse. What is the translation rate to human efficacy and safety in your historical data, and where have capsids that excelled in NHP failed in humans?
12. How critical is the Trisk Bio manufacturing partnership — does GMP capacity gate how fast partners can take your capsids to clinic, and is manufacturing a moat or a dependency?
13. What is your current headcount and burn, and how has the team changed since the 2021 ~50-person base through the AAV downturn?
14. The NVIDIA relationship — is it strategic optionality, a compute dependency, or a path to an eventual acquisition/standalone AI-bio business?
15. If you could only prove one thing to the market in the next 12 months to re-rate Dyno, what would it be — and what's blocking it?

---