Merge Labs

Phase A — Understand the business

Company Overview

Merge Labs is a pre-revenue, pre-clinical neurotechnology research lab building brain-computer interfaces (BCIs) that read and modulate brain activity without electrode implants, using ultrasound and engineered molecules instead of electrodes. It defines itself as a "research lab" with the mission of "bridging biological and artificial intelligence to maximize human ability, agency, and experience". There is no product, no clinical indication, no deployment timeline, and no revenue — neurofounders.co, reading Merge's own job postings, notes "the company has not outlined specific product targets, clinical indications, or deployment timelines" and shows "early emphasis on building internal capability around the core technical thesis before defining downstream applications or products".

The technical thesis (the whole company in one paragraph). Merge's stated belief is that "increasing bandwidth and brain coverage of BCIs by several orders of magnitude while making them much less invasive requires developing entirely new technologies that connect with neurons using molecules instead of electrodes [sonogenetics / acoustic reporter genes], and transmit and receive information using deep-reaching modalities like ultrasound". Two mechanisms underlie this (detailed in Lens 2/3):

• Reading — functional ultrasound imaging (fUSI): estimate neural activity by detecting blood-flow (hemodynamic) changes via the Doppler effect; ~100 µm spatial resolution, deep penetration, brain-wide field of view. It measures blood flow, not spikes — and it needs a skull window (the core caveat).
• Writing / modulation — focused ultrasound (FUS) neuromodulation: acoustic pressure opens mechanosensitive calcium ion channels to excite/inhibit neurons "with submillimeter accuracy"; this part does work through intact bone.
• The moonshot layer — sonogenetics: engineer target neurons (via gene delivery) to express acoustic reporters / gas vesicles so they become visible-to and controllable-by ultrasound — the "molecules instead of electrodes" line. This is Mikhail Shapiro's Caltech research program, and it is the part that, if it worked through the skull in humans, would be genuinely category-defining. It is also the part furthest from human validation.

Who the "customers" will be: undefined. The long-run framing is consumer-adjacent — Altman's pitch is a "think a thought, get a ChatGPT response" read-only, low-friction link to an AI agent (contrast Musk's "high-bandwidth symbiosis"). OpenAI's stated rationale is explicitly distribution-circular: "BCIs will create a natural, human-centered way for anyone to seamlessly interact with AI" — i.e., a new input layer for ChatGPT. There is no medical indication, no reimbursement path, and no SKU. n/a — private, not disclosed on every revenue line.

Contract / payment structure: none exists. Unlike an operating company, there is nothing to assess here — no recurring revenue, no take-or-pay, no concentration. The only "contract" of note is the OpenAI collaboration (scientific foundation models + frontier tools), terms undisclosed.

Leadership ambiguity (flagged, not resolved). Sources conflict on who actually runs Merge day-to-day. Several outlets call Alex Blania "CEO" of Merge Labs while simultaneously reporting he "will continue his role at Tools for Humanity" (Worldcoin), and Sumner Norman is described as CEO of Forest Neurotech. It is not clear from public sources whether Merge has a dedicated full-time CEO or is run by part-time founder-operators plus the Forest/Caltech scientific bench. I treat the executive-attention question as a live risk (Lens 9/13), not a settled fact.

Supply Chain (names or it didn't happen)

There is no commercial-layer file for the bci beat (supply-chain.md missing), so this is web-derived. Map of the value chain Merge sits in:

Upstream — the enabling hardware & science:

• Butterfly Network (NYSE: BFLY) — ultrasound-on-chip (CMUT semiconductor transducer). Forest Neurotech's Forest 1 device — a research-grade scanner "smaller than a standard key fob" — is "powered by Butterfly™ technology". This is the named, load-bearing hardware dependency: miniaturized, low-cost ultrasound on a chip is what makes an implantable/wearable fUS device conceivable at all. A real public-market read-through.
• Caltech / Shapiro Lab + Andersen Lab — the IP and talent source: gas-vesicle/acoustic-reporter sonogenetics (Shapiro) and closed-loop fUS motor decoding (Andersen/Norman/Griggs). The science is the supply chain at this stage.
• Forest Neurotech — the nonprofit Focused Research Organization (FRO) (Convergent Research model) that did the foundational human fUS work; Merge is effectively the for-profit commercialization vehicle of Forest's science. Forest "will maintain operations" with a "wonderful working relationship" with Merge.
• OpenAI — supplies "scientific foundation models and other frontier tools" — i.e., the AI to decode the (low-SNR, hemodynamic) ultrasound signal into intent. A genuine differentiator if the signal exists to decode.
• Gene-delivery / AAV supply chain (implied by the sonogenetic ambition) — any "molecules instead of electrodes" path requires viral or other gene delivery to neurons, pulling in the same CDMO/AAV manufacturing and FDA gene-therapy regulatory chain that gates the cell-and-gene-therapy industry. Unnamed by Merge, but structurally unavoidable for the moonshot layer.

Midstream — Merge Labs: integrates chip-ultrasound + decode-AI + (eventually) sonogenetic molecular reporters into a device/platform. No fab, no clinical sites of its own disclosed (Forest used UCLA Neuro-ICU and Cedars-Sinai).

Downstream — the buyer: undefined and unbuilt. Long-run intent is consumer/AI-interface (route to OpenAI/ChatGPT) rather than the medical-reimbursement channel that Neuralink/Synchron/Precision target. No distribution, no payer, no regulatory clearance.

Chokepoints / single-source dependencies: (1) Butterfly's ultrasound-on-chip is the named hardware chokepoint; (2) the skull is the physical chokepoint (see the grounding note — fUS recording can't cross it non-invasively today); (3) gene delivery is the chokepoint for the sonogenetic layer. All three are unresolved.

Competitive Advantages (moats)

What is genuinely differentiated:

1. A non-electrode modality with a real scientific pedigree. fUS is not vaporware — it has peer-reviewed human results (Forest's first human fUS imaging; the Andersen/Norman Nature Neuroscience 2023 closed-loop motor-decoding paper) and a credible physics story: ~100 µm resolution and brain-wide, deep coverage that electrode arrays (which sample a tiny cortical patch) cannot match. If the skull problem is solved, the coverage advantage over intracortical arrays is real and large.
2. Capital + brand + distribution moat (the Altman/OpenAI flywheel). A $250M seed at inception, OpenAI as lead and AI-compute partner, and a direct line to the largest consumer-AI surface on earth (ChatGPT) is a distribution and capital advantage no other BCI has. This is the actual moat today — it is a capital-and-narrative moat, not a technology moat.
3. Talent concentration. Shapiro (sonogenetics), Norman + Aflalo (fUS-BCI, ex-Caltech/Forest), Blania/Herbig (scaled Worldcoin's Orb hardware to 26M users) — a rare stack of ultrasound-neuroscience + consumer-hardware-at-scale operators.

Bargaining power: as a pre-product lab, Merge has none over a future supply chain or future customers — it has it only over talent and capital (where the Altman halo gives it leverage). Butterfly/gene-delivery vendors would hold the leverage in any real build.

Why the moat may be thinner than it looks (carried into Lens 13): the defensible asset is science that lives substantially at Caltech and Forest, with several founders part-time (Blania/Herbig at Tools for Humanity, Norman at Forest, Shapiro at Caltech). The IP is licensable/replicable by a well-funded rival (and the field is small — every serious fUS-BCI person knows every other one). The "moat" is money + Altman + OpenAI, which is formidable for fundraising and recruiting but does not protect a specific technical claim. Durability is unproven on every axis except capital access.

Segments

n/a — private, pre-revenue. There are no product segments, no geographic revenue split, no EBITDA — our figures is header-only and there is nothing to populate it with. The only meaningful "segmentation" is programmatic: a read program (fUSI decode), a write program (FUS neuromodulation), and a molecular program (sonogenetic reporters) — none of which is a revenue line. Recorded here as required; quantification is n/a — not disclosed.

---

Phase B — Measure performance (+private overlay: swap earnings/comps/projection for funding / cap-table / IPO-readiness)

Funding & Valuation Trajectory (swap for "Earnings Result")

All figures, unaudited.

That is the entire capital history — Merge is one round old.

The number that matters and why it's the story: ~$850M valuation at inception, pre-product, pre-clinical, pre-revenue. neurofounders.co: this "resembles late-stage financing in neurotechnology, despite Merge being newly formed," and "commits comparable-scale capital to a non-implant, experimental ultrasound-based approach at a point where the modality itself remains the primary uncertainty". For calibration: Neuralink took ~9 years and 11 implanted patients to reach ~$9B; Synchron is at ~$1B with 10 implants and an FDA IDE; Precision is ~$500M with an actual FDA 510(k) clearance. Merge is at $850M with zero patients, zero clearances, and an unproven recording modality — it is priced off founder + OpenAI optionality, not de-risking.

Burn signal: a research lab hiring "senior technical and research roles" to build ultrasound + AI-decode + (eventually) gene-delivery capability, in the Bay Area, burns on the order of **$30–60M/yr **. On ~$250M that is ~4–7 years of runway if it stays a lab and doesn't fund expensive human gene-therapy trials — generous by frontier standards, and the reason there's no urgency to define a product. The flip side: the first serious clinical/gene-delivery program would compress that fast.

No guidance, no print, no balance sheet to flag. our figures / our figures are header-only by necessity. There is nothing to reconcile because nothing has been reported.

Founder & Public Signal (sentiment) (swap earnings-call sentiment for founder interviews/announcements)

No earnings calls exist; the "sentiment" tape is announcements + founder interviews.

• Launch tone (Jan 2026): maximalist and mission-framed — "bridging biological and artificial intelligence," "several orders of magnitude" more bandwidth, "non-invasive." Notably omits Neuralink/Synchron by name in its own materials (a confident-or-evasive silence; see Lens 13).
• Altman framing: "think a thought, get a ChatGPT response" — a read-only, low-friction AI-input vision, deliberately more modest in interaction model than Musk's symbiosis pitch even as the valuation is aggressive.
• The recurring phrase to watch: "non-invasive." Every technical reviewer who looks closely flags that fUS recording isn't non-invasive yet — so the gap between what Merge says and what the science shows is the single most important sentiment signal here. When/if Merge stops saying "non-invasive" (or qualifies it to "minimally invasive"), that's a tell the company has met the physics.
• The Forest Neurotech body of work is the real signal of competence: "first-ever functional brain imaging in humans using ultrasound-on-chip," collaborations at UCLA Neuro-ICU and Cedars-Sinai, detection of auditory signals in TBI patients, and the open-sourced beamforming software ("an order-of-magnitude improvement on the state of the art"). This lineage is why the round is fundable.

Cap Table, Secondary Marks & Mechanism Comps (swap for P/E comps)

Syndicate quality. Lead is OpenAI (strategic, not a crossover fund — and the source of the conflict overhang, Lens 9/10), plus Bain Capital (real institutional money), Interface Fund, Fifty Years (deep-tech/longevity), and Gabe Newell (Valve founder, a known neurotech enthusiast/Starfish-adjacent). Tell: this is a strategic + deep-tech-VC + celebrity-angel syndicate, not a Fidelity / T. Rowe / Coatue crossover cap table — so the classic IPO-proximity signal is absent, exactly as you'd expect one round in. The Altman/OpenAI anchor substitutes brand for crossover validation.

Secondary marks: none disclosed. Forge / PitchBook / Tracxn list Merge but carry only the $850M primary mark (Jan 2026) — there is no secondary trading or markup/markdown to cite. n/a — not disclosed.

Mechanism / valuation comps — the BCI cohort (private marks,; never fabricated — n/a where unsourced; cross-checked against this shelf's peer dossiers):

The comp read (this is the whole valuation argument): Merge at $850M sits above Precision ($500M, FDA-cleared, shipping) and near Synchron's earlier rounds ($1B, 10 implants, FDA IDE) — with none of their clinical de-risking. On evidence-per-dollar it is the most expensive BCI in the cohort. The bull retort (Lens 12) is that it is the only one with a credible non-invasive-coverage path and OpenAI distribution; the bear retort (Lens 13) is that Kernel is the cautionary comp — a well-funded non-invasive-BCI dream that ran out of road when the signal proved too thin to build a business on. Merge is being priced like the bull case is already true.

Funding / Product-Event Catalysts (what moves the mark) (swap stock-catalysts for private value events)

No public stock, so "catalysts" are mark-resetting events:

• 15 Jan 2026 — stealth exit + $252M/$850M seed (OpenAI-led). The launch is the only realized value event; it set the mark off founder+OpenAI optionality.
• What would re-rate it up: first human fUS recording through (or substantially through) intact skull; first sonogenetic readout in a large animal/human; a named clinical indication + IRB pathway; a crossover-fund Series A (the missing IPO-proximity tell); deeper OpenAI productization (a ChatGPT input demo).
• What would re-rate it down: a credibility hit on "non-invasive" (a public technical rebuttal landing); the House Oversight / state-AG / SEC scrutiny of Altman's self-dealing forcing OpenAI to unwind or distance the investment (Lens 10); a Worldcoin-style regulatory/biometric blow-up tainting the founders' "scan your body for AI" brand; a rival (Neuralink/Synchron/an academic group) demonstrating non-invasive high-bandwidth first.
• Pattern read: because there is no tape and no product, the mark will move on (a) science milestones and (b) Altman/OpenAI headline risk — not on revenue. This is a headline-and-physics name, not a fundamentals name.

---

Phase C — Judge people & books

Management

Founders (the asset and the risk):

• Sam Altman (co-founder, board member, reportedly no equity) — the gravity well. Brings capital, OpenAI, narrative. Also brings the conflict-of-interest overhang (Lens 10) and is famously not an operator of this company day-to-day.
• Alex Blania (described as CEO; concurrently CEO of Tools for Humanity / Worldcoin) — track record: real and double-edged. He scaled Worldcoin's custom biometric Orb hardware to ~26M total users / 12M verified and raised >$300M at ~$2.5–3B for Tools for Humanity. That is genuine consumer-hardware-at-global-scale execution — exactly the muscle a BCI needs eventually. But the same playbook produced bans/investigations in Spain, Germany, Kenya (declared illegal, May 2025), Brazil, Indonesia and others over iris-biometric/GDPR violations. The pattern — scale first, ask regulators later, collect bodies' data for an AI future — is a flashing yellow light for a company that will literally read people's brains.
• Sumner Norman (co-founder; CEO of Forest Neurotech) — the strongest scientific operator here: pioneered ultrasound-based BCI at Caltech (Andersen lab), co-first-author of the Nature Neuroscience (2023) closed-loop fUS motor-decoding paper, took Forest "from lab to human patient in the shortest time ever achieved by comparable neurotech efforts".
• Tyson Aflalo (co-founder, ex-Caltech/Andersen lab) — established human intracortical/decoding researcher.
• Mikhail Shapiro (Caltech professor; remains at Caltech) — the sonogenetics intellectual engine: gas vesicles, acoustic reporter genes, mechanosensitive-channel ultrasonic neuromodulation; runs a 28-person Caltech lab.
• Sandro Herbig (product/eng; concurrently at Tools for Humanity) — Orb hardware/scaling.

Skin in the game / tenure: founders are brand-new to this company and several are explicitly part-time (Blania/Herbig at TfH, Norman at Forest, Shapiro at Caltech). Altman holds no equity (per reports) — which removes one conflict but also dilutes "skin in the game." The executive-attention question is the central management risk: who wakes up every day owning Merge's P&L of progress? Public sources do not answer this.

Capital-allocation history: too new to judge at Merge. The adjacent record (Worldcoin) shows founders who raise big and spend on global hardware scale-up aggressively — capital-intensive, regulation-courting. No value-destruction at Merge yet because nothing has been allocated beyond hiring.

Founder vs. professional manager: founder/visionary archetype, heavily — mission-first, narrative-first, science-bench-deep, operationally part-time at the top. Implication: fundraising and recruiting will be elite; focus, regulatory discipline, and accountability are the open questions.

Forensic Red Flags + Regulatory (re-pointed for private/pre-revenue)

Accounting/financial forensics: n/a — no financials exist. There is no income statement, balance sheet, or cash-flow statement to interrogate — no revenue recognition, receivables, SBC, or goodwill to flag. The honest forensic statement for a one-round-old private lab is: there is nothing to audit, and that opacity is itself the caveat — every figure in this dossier is company-stated or press-reported, none independently verified.

Regulatory findings (required sub-section). Read companies/merge-labs/regulatory/regulatory-findings.md (written 2026-06-29 by our model):

• SEC (EDGAR EFTS — LR + AAER): 0 findings. "Merge Labs has no CIK — it is private and not required to file with the SEC. No EDGAR enforcement search is possible.".
• Item 3 (Legal Proceedings): n/a — no 10-KForm 10-KA company’s audited annual report to the US regulator. The most complete thing it publishes. exists.
• Non-SEC web search ("Merge Labs" (FTC OR DOJ OR FDA OR...) enforcement): no enforcement action against Merge Labs itself found as of 2026-06-29 — it is too new and pre-clinical to have a regulatory record.
• Two indirect regulatory overhangs are material, however, and belong here:

1. The OpenAI/Altman self-dealing investigation. OpenAI's investment into its own CEO's startup is under active scrutiny: the U.S. House Oversight Committee launched an investigation (May 2026), and six Republican state attorneys general (FL, MT, NE, IA, WV, LA) urged the SEC to review OpenAI's governance before any IPO; OpenAI responded by "strengthening its Conflict of Interest Policy" and adding a whistleblower hotline. Altman "sits on the Merge Labs board but reportedly holds no equity." This is a governance/regulatory cloud over Merge's anchor investor and could force a distancing.
2. The founders' Worldcoin regulatory record (Blania/Herbig): iris-biometric operations banned/suspended/declared illegal across Spain, Germany, Kenya, Brazil, Indonesia and more for GDPR/consent violations. Not Merge's record — but the same operators applying a "collect human biometric/neural data for an AI future" model, which regulators have already shown they will block. A direct read-through to how a brain-reading consumer product would be received.

• Net: "No SEC/EDGAR record (private, no CIK); no direct enforcement action against Merge Labs found via web search as of 2026-06-29; but the anchor investor (OpenAI/Altman) is under House Oversight + state-AG/SEC governance scrutiny, and the founding operators carry a documented multi-country biometric-privacy ban history (Worldcoin)." Verified via SEC EDGAR EFTS (0), web search, and the absence of any company filing.

---

Phase D — Project & stress-test

IPO-Readiness & Path-to-Tradeable (swap forward-EPS for the be-early lens)

There is no EPS to project — pre-revenue, pre-product. No our model EPS line is logged (correct per --watchlist rules; there is nothing committable yet). The relevant projection is path-to-tradeable:

• Stage: seed / inception — the earliest possible. On the private-watch.json readiness scale (1=early/seed … 5=S-1 filed), Merge is a 1/5.
• IPO window: none in sight — 5+ years at minimum, contingent on a science breakthrough that hasn't happened. Before an S-1 is even imaginable Merge must: (a) demonstrate human brain recording that doesn't require a craniectomy (the physics gate); (b) define a clinical indication or a shippable consumer product; (c) clear FDA (device, and gene-therapy review if sonogenetics is in the path — a multi-year process on its own); (d) attract a crossover round (the missing IPO-proximity tell). Realistically this is a late-2020s-to-2030s question, if ever.
• More-likely outcomes than IPO: (1) continued private mega-rounds funded by the Altman/OpenAI orbit (the path of least resistance — it can stay private and well-capitalized for years); (2) absorption into OpenAI (the circular logic ends in acquisition if BCI becomes a strategic input layer); (3) quiet wind-down or pivot if through-skull fUS proves intractable — the Kernel precedent (non-invasive BCI → wellness pivot → capital starvation) is the live cautionary base case for a non-invasive bet.
• Be-early payoff / write-back: Merge Labs should be added to research/private-watch.json under beat: bci with stage: seed, ipo_readiness: 1, lead OpenAI (+ Bain, Fifty Years, Gabe Newell), catalyst "$850M-at-inception ultrasound/sonogenetic non-invasive BCI; modality unproven through skull; OpenAI conflict overhang", and dossier set to this file. (Per wave boundaries I do NOT edit that file here — flagged for Connor's our position log/coverage pass.)

The forecast that is worth tracking (binary, not EPS — log later if conviction forms): "Merge Labs publicly demonstrates human brain-activity recording (fUS or sonogenetic) without skull removal / cranial window by YE2028." My prior : low — ~20–30% given that Forest's own science still needs a window. That single binary, not a revenue number, is what should be calibrated for this name.

Bull vs Bear

Bull case. Merge is the only BCI attacking the prize that actually matters for a consumer AI-interface era: whole-brain, deep, non-electrode coverage at scale, with a path to non-invasive. Electrode arrays (Neuralink/Precision) sample a postage-stamp of cortex and require neurosurgery — a ceiling on TAM. fUS already has ~100 µm resolution, deep penetration, and peer-reviewed human results; pair that with OpenAI's decode models and Butterfly's ultrasound-on-chip miniaturization, and you have a credible 10-year route to a wearable, no-surgery brain input for ChatGPT — a market measured in hundreds of millions of consumers, not thousands of paralysis patients. The founders combine the best fUS-BCI scientists alive (Norman, Aflalo, Shapiro) with operators who actually scaled global consumer hardware (Blania/Herbig). With Altman + OpenAI + $250M, it has the capital, brand, and distribution moat to out-recruit and out-fund every rival. At $850M it is "cheap" relative to a Neuralink at ~$9B if the non-invasive thesis lands — because non-invasive is worth an order of magnitude more TAM than invasive. Contrarian view the market is refusing to see (bull side): the BCI race is being scored on bandwidth and implants (Neuralink's frame), but the consumer-AI era will be won on zero-surgery + good-enough read-only bandwidth — and Merge is the only well-capitalized team building exactly that.

Bear case (the stronger case, given the price). Three things could permanently impair this:

1. The physics may not yield. fUS recording does not work through the intact skull — Merge's own scientific lineage (Forest) only got human data in craniectomy patients, and the field "has tended to move incrementally… rather than directly toward general-purpose neural interfaces". If through-skull fUS is a 15-year problem (or impossible at useful SNR), the entire non-invasive premise collapses to "another implant company, but later and worse than the incumbents."
2. The sonogenetic moonshot adds a second miracle. "Molecules instead of electrodes" means gene-delivering acoustic reporters into human neurons — stacking the entire gene-therapy safety/FDA/manufacturing risk surface on top of the ultrasound risk. Two unproven hard things in series multiply, not add.
3. It's priced for the bull case to be already true. $850M at inception, above an FDA-cleared Precision, demands flawless execution on an unproven modality — and the Kernel precedent shows exactly how a well-funded non-invasive BCI dies (signal too thin → wellness pivot → starvation).

Pre-mortem (18 months out, thesis broke): it's mid-2027. Merge has published nothing showing through-skull human recording; the work that exists still needs a window. The House Oversight / SEC scrutiny of Altman's self-dealing has made OpenAI quietly distance itself, removing the distribution-and-capital halo. A Worldcoin-style biometric backlash has attached to the founders' "scan-you-for-AI" brand, poisoning the consumer narrative. Meanwhile Synchron/Precision keep shipping real clinical milestones. The $850M mark looks like a founder-premium top-tick, and the next round is a flat-or-down strategic check from the Altman orbit because no crossover will touch it. What happened: the market paid for the story before the physics paid out.

Are multiples too high? There are no multiples (no revenue). On evidence-per-dollar, yes — it is the most expensive, least de-risked name in the BCI cohort. The valuation is pure option value on a science breakthrough + an Altman/OpenAI exit.

Devil's Advocate (short-seller)

Dismantling the bull case:

• The structural break in how it "makes money": it doesn't, and the path to doing so requires solving a problem its own founders haven't solved. Strip the Altman halo and Merge is a seed-stage lab with no product, no indication, no clinical site of its own, and a recording modality that needs a hole in the skull — valued at $850M. The "non-invasive" claim, the load-bearing differentiator, is contradicted by the company's own scientific lineage.
• Where the value is concentrated, and the fragility: the value is Sam Altman + OpenAI, not Merge's IP. That anchor is under active government scrutiny for being a conflict of interest (House Oversight, six state AGs, SEC asks). If OpenAI is forced to unwind or distance the investment — or if Altman's standing wobbles — the thesis loses its capital, brand, and distribution in one move. Single-point-of-failure risk dressed up as a moat.
• The moat is weaker than bulls think because the science is licensable and lives at Caltech/Forest, the founders are part-time at the top, and the field is tiny. A rival with capital can hire the next Shapiro student. The only thing not replicable is Altman himself — which is exactly the asset under regulatory cloud.
• The most dangerous competitor bulls underestimate: not Neuralink — Kernel (and the category of non-invasive BCI). Kernel had the non-invasive dream, real funding, and a serious founder, and the non-invasive signal proved too thin to build a business → wellness pivot → no priced round since 2020 [companies/kernel]. Merge is a better-funded, better-branded re-run of the same bet, and the physics hasn't changed. Separately, Synchron quietly de-risks the "minimally-invasive + good-enough-bandwidth + Apple-distribution" lane that could satisfy the consumer-input use case years before Merge's through-skull dream.
• Worst capital-allocation / governance flags: the founders' Worldcoin playbook (collect global biometric data, court regulators after the fact, eat multi-country bans) applied to neural data; Altman's self-dealing (OpenAI funding his own startup) as the company's origin; part-time founder-operators with divided focus; and maximal valuation with minimal disclosure ("public information remains limited to high-level statements").
• What must hold for $850M: (1) through-skull human fUS recording becomes real; (2) it does so at bandwidth useful for AI interaction; (3) FDA (and gene-therapy review, if sonogenetic) clears a consumer or clinical path; (4) OpenAI/Altman stays the anchor through the conflict probe; (5) a crossover round validates the mark. If any one fails, the valuation is impaired.
• Growth disappoints by 20–30%? There is no growth to disappoint — the equivalent stress is a failed/null through-skull readout, which doesn't haircut the valuation, it breaks the thesis (binary, not linear).
• Single scenario that permanently impairs it: through-skull fUS recording proves infeasible at useful SNR in humans (the most plausible bad outcome, given Forest's own craniectomy dependence). Plausibility: moderate-to-high — this is the base-rate path for non-invasive BCI, not a tail.

Management Questions (ordered by information value)

1. Has Merge — or anyone — recorded usable human brain activity via functional ultrasound through the intact skull? If not, what is the specific physics result that makes you believe it's achievable, and by when? (The entire thesis lives or dies here.)
2. Is "non-invasive" your shipping claim, or an aspiration? At launch, what is the minimum invasiveness (cranial window? thinned skull? full transcranial?) of your first human recording system?
3. Does the sonogenetic/acoustic-reporter path require gene delivery into human neurons? If so, how do you carry the FDA gene-therapy risk surface, and is that in the first product or a later one?
4. Who is the full-time CEO and executive team of Merge Labs? Which founders are full-time vs. retaining roles at Tools for Humanity / Forest / Caltech, and how is day-to-day accountability structured?
5. What is your first defined product — consumer AI-input or clinical/medical — and what is the indication and regulatory pathway?
6. How is the OpenAI relationship governed given the conflict of interest? What did the OpenAI board approve, who recused, and what protects Merge's independence (and OpenAI's shareholders) — especially under the current House Oversight / SEC scrutiny?
7. What does Sam Altman's role entail operationally, and is it accurate that he holds no equity? How is that incentive structure intended to work?
8. What is the relationship and IP boundary between Merge (for-profit) and Forest Neurotech (nonprofit FRO)? Who owns the foundational fUS and beamforming IP?
9. What is your bandwidth target (bits/sec or effective WPM) for a first human read-only interface, and how does that compare to what Forest's craniectomy data achieved?
10. Given the founders' Worldcoin regulatory history, how will you handle consent, data ownership, and global privacy regulation for neural data from the outset?
11. What is your current runway and burn, and what milestone does the $252M seed fund to before the next raise?
12. fUS measures hemodynamics (blood flow), not spikes — what is the latency and temporal-resolution ceiling that implies for real-time AI interaction, and how do OpenAI's models compensate?
13. Beyond Butterfly Network, what is your hardware supply chain, and how do you de-risk single-source dependence on ultrasound-on-chip?
14. What would falsify your approach — what result in the next 24 months would make you abandon through-skull fUS?
15. Why is $850M the right valuation for a pre-clinical lab, and what does that price assume the market currently misunderstands about non-invasive BCI?

---