Sanctuary AI

Phase A — Understand the business

Lens 1 · Company Overview

Sanctuary AI (legal: Sanctuary Cognitive Systems Corporation) builds Phoenix, a general-purpose bipedal humanoid robot, and Carbon, the AI control system that drives it. Founded 2018 in Vancouver by Geordie Rose (ex-D-Wave, ex-Kindred) and Suzanne Gildert. The original mission was maximalist — "the world's first human-like intelligence in general-purpose robots," an AGI-in-a-body moonshot. Carbon was pitched as a reasoning/task/motion-planning stack whose plans are "explainable and auditable," trained on an "alphabet of movement" (touch, grasp, push, rotate, place).

What it actually sells today (the post-pivot reality): after a 2024 leadership purge, the company narrowed from "autonomous AGI humanoid" to its one genuinely differentiated layer — dexterous robotic hands plus teleoperation-assisted manipulation. The stated commercial model is "labor-as-a-service": customers pay a negotiated rate for completed work, no capex, no line changes. No public price card; humanoid pilot units in this class run ~$100K–$250K of hardware value.

• Products/services: Phoenix (8 generations since 2022; Gen 7 = April 2024, <24h task learning; Gen 8 = Jan 2025, optimized for data capture); Carbon control stack; standalone dexterous-hand IP (21 DOF/hand, hydraulic, tactile to 5 mN).
• Main customers / pilots: Magna International (automotive sub-assembly, multi-unit deployment); Canadian Tire / Mark's / Sport Chek (retail pilots — inventory, folding, labeling, "100+ in-store tasks"); Microsoft (go-to-market accompaniment at Hannover Messe 2025; ambition stated as "deploy a thousand robots").
• Suppliers: Magna doubles as a strategic manufacturing/cost-down partner (its automotive parts portfolio applied to Phoenix BOM); otherwise the actuator/sensor stack is largely in-house (the hydraulic hand is proprietary).
• Competitors: Figure, Tesla Optimus, Apptronik, 1X, Agility, Boston Dynamics, Unitree (see Lens 3 / Lens 7).
• Contract structure: pilots, not committed take-or-pay volume. No disclosed recurring revenue or backlog. Concentration is effectively total — a handful of unpaid/low-paid pilots, no at-scale paying fleet.

Plain-terms read: Sanctuary is a deep-tech hardware-IP company that mistook itself for an AGI lab for six years, and in 2024 was forced by its board to become a manipulation-component business. The hands are real and best-in-class. The "company that ships autonomous humanoid labor at scale" is not — it's a prototype-and-pilot operation running on fumes.

Lens 2 · Supply Chain

Map: upstream inputs → Sanctuary → end customer.

• Upstream (named):
• Hydraulic actuation / hands — proprietary, designed in-house; this is the crown-jewel sub-assembly (21 DOF, fluidic). Sanctuary and Clone Robotics are the two names pushing fluidic/hydraulic actuation against the industry's electric-motor default.
• Tactile sensors — new-generation finger-pad sensors integrated Feb 2025, sensitivity 5 mN; supplier not disclosed (likely in-house or specialist).
• Compute / AI — Carbon stack in-house; Microsoft is the cloud/AI go-to-market partner (Azure ecosystem implied).
• Manufacturing cost-down — Magna International explicitly engaged to bring automotive-grade manufacturing and parts to lower Phoenix BOM and improve scalability.
• Midstream (Sanctuary): integrates hands + body + Carbon into Phoenix; assembly is hand-built / low-volume ("likely a hand-built prototype," no mass line evidence).
• Downstream (named buyers): Magna (factories), Canadian Tire group (retail), Microsoft (channel), with stated-but-unproven ambition of 1,000-unit deployments.

Chokepoints / single-source dependencies:

1. Capital is the binding chokepoint, not silicon. Multiple analysts state plainly: "the biggest risks are financial and organizational, not technical". The supply chain is starved of the one input — money — that lets the rest scale.
2. In-house hydraulic hand = single-source on its own differentiator. If that team is poached (and the founder/CTO have already left), the moat walks out the door.
3. No at-scale manufacturing partner committed. Magna is a pilot/cost-study relationship, not a contract-manufacturing volume commitment.

This lens fails if it stays generic — names delivered: Magna, Microsoft, Clone (peer), Canadian Tire / Mark's / Sport Chek, plus the in-house hand/sensor stack. The honest verdict: a thin chain with no committed volume node.

Lens 3 · Competitive Advantages (moats)

The one real moat: intellectual property on hands + control.

• #3 globally in humanoid/embodied-AI patents per Morgan Stanley's published-US-patent ranking; the only startup in the top-20 patent holders alongside incumbents.
• 77 patent families, 277 publications, 7 jurisdictions (US, Canada, EPO, WIPO, Japan, China, Australia). Tech+IP Advisory: among humanoid startups, "only Sanctuary AI and Brain Corp make the list".
• Hardware lead in dexterity: 21 DOF/hand, hydraulic actuation, tactile to 5 mN — described as "far ahead of any competitor's hand design," with "beyond-human-capability" kinematics.

The moat that doesn't exist: everything that needs capital.

• No fleet-data flywheel at scale — the data moat in this category accrues to whoever has the most robots doing the most real work (Figure/BMW: 30,000+ vehicles, 11 months on-line; Apptronik/Mercedes+GXO; Tesla's internal fleet). Sanctuary has pilots, not a fleet.
• No manufacturing scale / vertical integration vs Tesla and Unitree.
• No balance-sheet moat — the defining weakness (Lens 5).

Bargaining power: weak on both sides. Against customers (Magna, Microsoft) Sanctuary is the supplicant — these are pilots the partners can walk from. Against the talent market it is losing — founder + CTO gone, ~30 laid off. The only asset with genuine bargaining power is the patent estate itself, which is precisely why the realistic exit is an IP sale/acqui-hire.

Durable-moat scorecard: IP/process = strong; switching costs = weak (pilots); network/data effects = weak (no fleet); scale = weak; brand = moderate (high-credibility pedigree, strong press). Net: a narrow but real moat on a company that may not survive to monetize it.

Lens 4 · Segments

No segment financials exist — private, segments.csv empty. By product/use-case the de-facto segmentation is:

By geography: HQ Vancouver; pilots in Canada (Magna, Canadian Tire) with stated US-expansion intent tied to the 2025 raise. No revenue split is disclosable — n/a — private, not disclosed.

Trend: the center of gravity shifted decisively from "full autonomous humanoid" → "hands + teleop-assisted manipulation" across 2024–25. That is a deceleration in ambition (good capital discipline, bad narrative for a venture priced on AGI dreams) forced by the board, not chosen from strength.

---

Phase B — Measure performance

Lens 5 · Funding & Valuation Trajectory (+private swap for "Earnings Result")

There are no earnings. The scoreboard is the cap table — and it tells a deteriorating story.

Round history (best-available, unaudited):

• Series A — C$75.5M (US$58.5M), 2021, oversubscribed; investors: Bell/BCE, Evok Innovations, Export Development Canada, Magna, SE Health, Verizon Ventures, Workday Ventures, DNX, Zeon, Harbrook.
• Strategic Innovation Fund (Government of Canada): ~C$30M non-dilutive.
• July 2024 round brought cumulative funding to ~US$140M; valuation US$221–232M (July 2024).
• Jan 2025 — US$10M convertible note to "finance the company for fiscal 2025," 2-yr term, converts at a 20% discount to the next priced round; only ~50% subscribed at announcement, existing holders asked to fill the rest. A convertible bridge at a discount, half-subscribed, is a textbook distress signal.
• April 2025 — attempting a US$175M "Series B-2" by selling its majority Apptronik stake (see Lens 7) and lining up NEA + Kleiner Perkins as US$10M co-leads; targeting another ~US$50M VC, minimum US$30M, end-May close.
• 2026 — total funding reported ~US$130–148.6M across sources; latest tracked event a tiny Series A-VI US$1.55M (2026-04-17). That cumulative total barely moving from the 2024 ~$140M figure implies the US$175M mega-round did NOT close at target — the company is grinding on small increments and asset sales, not a clean priced up-round.

Burn signals: $10M emergency bridge, half-subscribed; sale of the single best balance-sheet asset to fund operations; ~30 layoffs alongside the CEO ouster; "minimum target US$30M" language. Cash runway is the entire story and it is short..

Market reaction equivalent: the private market's verdict is a down-trajectory / stalled mark — note one source still tags the valuation at $221–232M (mid-2024) with no credible step-up since, while every major peer re-rated up 10–100x (Lens 7). In a sector that minted a $39B (Figure) and a $5B (Apptronik) name, a flat-to-down ~$230M mark is a relative collapse.

Lens 6 · Founder / Leadership Voice (sentiment trend) (+private swap for "Earnings Calls")

No earnings calls. Track the public voice instead — and it has been gutted.

• 2023–early 2024 — Geordie Rose, full visionary register: "There are problems facing humanity we can't solve — our robots could offer a solution". Carbon framed as a path to human-like cognition. Maximum ambition, AGI-adjacent.
• April 2024 — CTO Suzanne Gildert departs. Co-founder scientific voice gone.
• Nov 2024 — Geordie Rose ousted ("sudden, unexpected," board-forced) + ~30 layoffs; James Wells (CCO) → interim CEO. Tone resets from moonshot to "commercial pragmatism" — deploy, generate revenue, survive.
• 2025–26 — Olivia Norton surfaces as CTO in investment communications, signaling continued executive reshuffling.

Sentiment shift: from soaring (AGI/humanity) → defensive/operational (deploy, cut burn, survive) in roughly twelve months. The phrases they stopped saying — "human-like intelligence," "AGI" — are as telling as anything they added. The recurring new theme is teleoperation as a feature, not an embarrassment (a pre-emptive defense against the "your demos are puppeteered" critique; see Lens 13).

Lens 7 · Cap Table & Secondary Marks (+private swap for "Comps")

Cap-table quality: strategics-heavy, light on tier-1 crossover capital — and that is the tell. Investors include Bell/BCE, Magna, Verizon Ventures, Workday Ventures, Evok, Export Development Canada, BDC Capital (Thrive Venture Fund), InBC, SE Health. NEA + Kleiner Perkins appear only as proposed US$10M co-leads in the unclosed 2025 round. There is no Fidelity / T. Rowe / Coatue crossover entry — the classic IPO-proximity signal is absent. The syndicate skews Canadian strategic + government, which funds survival, not a $5B re-rate.

The defining cap-table event — Sanctuary sold its competitor stake to survive:

• Sanctuary acquired a majority stake in Apptronik in 2022 for US$10M.
• By 2025 that stake was worth ~US$125M at Apptronik's US$1.4B pre-money — a >12x.
• Sanctuary is selling it to fund its own operations. By Feb 2026 Apptronik raised $520M at a ~$5B valuation ($935M total). Sanctuary incubated, then liquidated early, the company now worth ~20x its own last mark. This is the single most damning fact in the file.

Peer marks (private valuations / scale) — provenance-critical, all ``:

Read: Sanctuary trades at a rounding-error valuation versus the cohort it helped define, with the strongest patent estate and the weakest balance sheet in the group. Multiples like EV/Sales are n/a — no revenue disclosed. The honest comp is distance-to-zero vs distance-to-IPO: peers are racing to scale; Sanctuary is racing the clock.

Lens 8 · Funding / Product Catalysts (events that moved the story) (+private swap for "Stock-Price Catalysts")

No stock, so track the value-moving events (and what they reveal about what actually matters for this name):

• May 2023 — Phoenix unveiled / "labor-as-a-service" → established the category narrative & press profile. Up.
• April 2024 — Magna strategic deployment + Gen 7 → first credible industrial validation. Up.
• April 2024 — CTO Gildert departs → first crack in the founding team. Down.
• July 2024 — funding to ~$140M, ~$230M valuation → last clean financing. Flat-up.
• Nov 2024 — Rose ousted + ~30 layoffs → the regime change; AGI-moonshot → survival-mode. Sharply down.
• Jan 2025 — $10M discounted convertible, half-subscribed → public distress signal. Down.
• Feb 2025 — new tactile sensors / dexterity milestones → reinforced the one real moat. Modest up (technical).
• April 2025 — Apptronik-stake sale to chase $175M → the company monetizes its best asset to survive; round appears not to have closed at target. Down (structurally).
• Hannover Messe 2025 — Microsoft GTM → channel credibility, ambition reset ("1,000 robots"). Up (narrative).

Pattern: the market for this name reacts to (1) capital/runway events and (2) team integrity far more than to technical demos. Dexterity milestones barely move the needle; a half-subscribed bridge and a founder ouster move it a lot. For Sanctuary, the dominant variable is solvency, not capability.

---

Phase C — Judge people & books

Lens 9 · Management

Founder archetype — visionary, serial, science-first, and now gone.

• Geordie Rose (co-founder, ex-CEO): PhD theoretical physics (UBC, 2000); founded D-Wave (1999, first company to sell quantum computers — Google, NASA, Lockheed, USG), founding CEO of Kindred (first RL-in-production robotics), then Sanctuary. Raised ~$1.7B across science ventures; inventor on 70+ US patents; Foreign Policy Top-100 Global Thinker (2013). Two of his companies hit MIT Tech Review's smartest-companies list (D-Wave #40 2014, Kindred #29 2017).
• The catch: Rose's pattern is pioneering category creation, not durable commercial scale. D-Wave's quantum value-delivery was perennially contested; Kindred was acqui-exited; Sanctuary burned six years on AGI ambition before the board forced a commercial reset. He is a world-class zero-to-one scientist and a questionable one-to-N operator — and in Nov 2024 the board removed him. He is now a strategic advisor at ExperienceFlow.AI — i.e., moved on.
• Suzanne Gildert (co-founder, ex-CTO): scientific co-founder; departed April 2024. The founding scientific brain-trust is fully gone.
• James Wells (interim CEO from Nov 2024): ex-CCO, ~5 yrs at Sanctuary; the "commercial pragmatist" installed to deploy and monetize.
• Olivia Norton (CTO, surfaced 2026): appears in 2026 investment comms — continued reshuffling.
• Philip Smith (CFO): ran the Jan-2025 convertible note process.

Track record (as a company): category leadership in IP and hand dexterity — genuinely best-in-class. But capital allocation is the indictment: the most consequential decision the company made — buying into Apptronik for $10M (2022) — was brilliant; selling it under duress to fund payroll, just before Apptronik 5x'd again, converted a generational asset into a bridge loan. That is value destruction by timing forced by undercapitalization.

Tenure & skin in the game: founders departed; insider ownership now diffuse across strategics/government. Alignment is weakened — the people who built the IP no longer run or (fully) own the company.

Red flags: board-forced CEO removal; co-founder/CTO exits; ~30 layoffs; emergency discounted convertible; fire-sale of the best asset; executive-suite churn (CEO and CTO both turned over inside ~18 months). This is a management table in active crisis-stabilization, not a confident operator on offense.

Lens 10 · Forensic Red Flags

No audited financials exist — so the forensic posture inverts: the risk isn't aggressive revenue recognition (there's little revenue to recognize), it's going-concern and capital-structure risk.

• Going concern (the headline): a half-subscribed $10M convertible note explicitly to "finance the company for fiscal 2025," a fire-sale of the Apptronik stake, and a $175M raise that appears not to have closed at target (cumulative funding barely moved 2024→2026) together constitute a flashing going-concern signal. Unaudited per public sources.
• Convertible-note overhang: the note converts at a 20% discount to the next priced round — structural dilution baked in, and a down-round trigger that punishes existing holders.
• Asset stripping to fund opex: selling the appreciating Apptronik position (not a core-business divestiture for strategic reasons, but a liquidity-driven sale) is the cash-flow-vs-value divergence equivalent — burning the balance sheet's best line to cover the burn.
• Valuation staleness: carrying a mid-2024 ~$230M mark with no credible step-up while peers re-rated is itself a soft red flag on mark-to-reality.
• Demo/disclosure integrity: persistent third-party claims that "autonomous" demo footage is substantially teleoperated — a narrative-integrity flag common to the sector but acute here given the pivot literally toward teleoperation.

Regulatory findings (required sub-section). Read from regulatory/regulatory-findings.md (pre-fetched Stage 1, 2026-06-18):

• SEC (EDGAR EFTS — LR + AAER): 0 findings. Sanctuary has no CIK — it is private and not required to file with the SEC; no EDGAR enforcement search is possible.
• Non-SEC enforcement (web search): ran "Sanctuary AI" (FTC OR DOJ OR FDA OR CFPB OR "consent decree" OR settlement OR fine OR penalty) enforcement — no material regulatory enforcement hits surfaced. Coverage is dominated by funding, leadership, and product news, not legal/regulatory actions.
• Item 3 (Legal Proceedings): n/a — private, no 10-K exists.
• Conclusion: No material regulatory or legal findings — verified via SEC EDGAR EFTS (LR, AAER, 0 hits), web search (no material hits), and confirmed-no-10-K, as of 2026-06-18. The material risk here is financial/going-concern, not regulatory.

---

Phase D — Project & stress-test

Lens 11 · IPO-Readiness & Path-to-Tradeable (+private swap for "Forward Projection")

Sanctuary is NOT in research/private-watch.json — so there is no pre-set stage/readiness to read; this lens assigns one from the evidence (per wave boundaries, this dossier does not write back to that file — flagged as an add candidate for Connor).

Assigned IPO-readiness: 1 / 5 (early/distressed) on the doc's scale (1=early/seed … 5=S-1/imminent).

• Stage: distressed-growth — post-pivot, post-purge, sub-scale, runway-constrained.
• Why a 1, not the 3 its robotics-peer cohort carries: Figure/Apptronik/Agility sit at 3–4 (late, mega-rounds, crossover-adjacent). Sanctuary has the opposite profile of an IPO-approaching company: stalled mark, asset sales, emergency bridge, no crossover capital, leadership turnover. An S-1 is not on the horizon.
• Milestones that would unlock a tradeable event (none imminent):

1. A clean, priced up-round at a stepped-up valuation with a tier-1 crossover lead → would reset the entire thesis. Has not happened despite a year of trying.
2. Paying, at-scale fleet revenue (not pilots) — a real labor-as-a-service contract with committed volume.
3. Manufacturing scale at a BOM that pencils against the ~$10–15/hr labor-payback threshold.

Most-probable path-to-tradeable — and it isn't an IPO: acqui-hire or IP/patent-portfolio sale. The #3 humanoid patent estate + best-in-class hands are acquisition-grade assets; the company is survival-grade. A larger, capital-rich player (a hyperscaler, an auto OEM, a Figure/Tesla/Apptronik, or a Chinese major) buying the patents + hand team is the base case for how shareholders realize value.

Forecast (NOT logged — --watchlist/private; no forecast.ts create):

• Base: Sanctuary does not complete an independent IPO within 3 years; resolution is an M&A/IP sale (~55–65% likelihood) or a heavily dilutive distressed recap (~20–25%), with outright wind-down a real tail (~15–20%). Inputs: stalled $230M mark, failed $175M round, $10M bridge, asset sales, founder exits — all ``.
• Bull: a strategic acquirer or crossover pays up for the IP and recapitalizes the hand/Carbon program inside a balance sheet that can actually fund scale.
• Bear: the cash runs out before a buyer agrees on price; the estate sells for scrap in a distressed process and equity holders (esp. below the convertible) are impaired.

Lens 12 · Bull vs Bear

Bull case. Sanctuary owns the single most defensible asset in humanoid robotics that money can't quickly replicate: the patent estate (#3 globally, 77 families) and a genuinely beyond-human dexterous hand (21 DOF, hydraulic, 5 mN tactile). In a field where manipulation is the unsolved problem and everyone else is racing on locomotion and whole-body AI, Sanctuary is deepest exactly where the value is hardest to reach. If a capital-rich acquirer drops this IP into a balance sheet that can fund manufacturing scale, the hands could become the de-facto manipulation layer of the industry — a Mobileye-style component-IP outcome. The Magna/Microsoft relationships and the labor-as-a-service framing give a credible commercial wedge. Pre-mortem-proof version of the bull: you're not buying a humanoid OEM, you're buying the hand-and-control IP at a ~$230M mark that a $5–39B peer could justify acquiring tomorrow.

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

1. Insolvency before monetization. The most likely cause of permanent impairment isn't competition — it's running out of cash. A half-subscribed bridge + a failed mega-round + a fire-sale of the best asset is the profile of a company that may not see 2027 intact.
2. Talent flight evaporates the moat. Patents protect filings, not people. With the founder and CTO already gone and ~30 laid off, the hand/Carbon team is poachable — and in this market, Figure/Apptronik/Tesla can outbid for exactly those engineers. A patent estate with no team to extend it is a depreciating asset.
3. The category out-scales them on a different axis. Even if the hands are best-in-class, the winners are being decided by fleet-data + manufacturing scale (Figure's 30k-vehicle BMW run, Tesla's 50k/yr line, Unitree's cost curve). Sanctuary could be technically right about manipulation and commercially irrelevant because it can't field enough robots to matter.

Pre-mortem (18 months out, thesis broke): The $175M round never fully closed; a second bridge diluted everyone; NEA/Kleiner walked when scale revenue didn't appear; a strategic acquirer lowballed the patent estate in a distressed process; the hand team was hired away by a $5B+ peer before the deal closed. The company is remembered as "the one that invented the best hands and sold Apptronik to make rent."

Are the multiples too high? Paradoxically, no — at ~$230M the equity may be cheap relative to the IP for an acquirer. But for a standalone-going-concern thesis the price is irrelevant because the going concern is in doubt. It's cheap as an asset and uninvestable as a company.

Contrarian view the market is refusing to see: consensus treats Sanctuary as a fading also-ran in a two-horse (Figure/Tesla) race. The thing being missed: Sanctuary's patent estate is a strategic landmine for every well-funded competitor. Whoever acquires it gets both the hands and a freedom-to-operate cudgel against rivals. The under-priced scenario isn't "Sanctuary wins" — it's "Sanctuary's corpse gets fought over," and the IP fetches more than the current equity mark implies.

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

Dismantling the bull case as a skeptical short-seller would:

• The moat is a museum piece. Patents on hydraulic hands matter only if hydraulic wins — and the entire rest of the industry (Figure, Tesla, Apptronik, 1X) is going electric for serviceability, cost, and density. Sanctuary may hold a commanding IP position in a road not taken. "Beyond-human" 21-DOF dexterity is over-engineering for picking parts in a Magna plant — customers want reliable, cheap, and serviceable, not anatomically transcendent.
• Revenue is concentrated at ~zero. There is no disclosed paying fleet — pilots that partners can cancel costlessly. If Magna or Microsoft walks, there's nothing underneath.
• The most dangerous competitor bulls underestimate: Apptronik — the company Sanctuary itself incubated. It went from a $10M Sanctuary investment to $5B and $935M raised while Sanctuary stalled. The student is now 20x the teacher, and Sanctuary sold the position. Also China (Unitree et al.) on cost — $20–50K Chinese humanoids reset the price floor.
• Worst capital-allocation move: selling the Apptronik stake under liquidity duress, just before another 4–5x — turning the company's one home-run into emergency cash.
• Assumptions that must hold for any equity value: (a) the company closes a real round soon; (b) the hand team stays; (c) a buyer values patents richly. All three are shaky.
• If growth disappoints 20–30%: there is no "growth" to disappoint — the relevant shock is runway, and a single failed financing window is terminal, not a haircut.
• Single scenario that permanently impairs: cash exhaustion → distressed asset sale below the convertible's preference → common equity zeroed. Given the disclosed financing stress, this is plausible, not tail.

Short-seller's one-liner: Best hands in the business, bolted to a going-concern problem; the IP is worth more than the company, which is the whole problem.

Lens 14 · Management Questions (ordered by information value)

1. What is current cash runway in months at present burn, and what is the hard date by which a financing must close? (The thesis-deciding question.)
2. Did the April-2025 ~US$175M round close, and at what valuation and structure — up, flat, or down vs the ~$230M mid-2024 mark?
3. What were the net proceeds from the Apptronik stake sale, and how much runway did they buy?
4. How much paying, contracted (non-pilot) revenue exists today, and what is committed backlog with Magna / any customer?
5. On your "autonomous" deployments, what fraction of task-time is truly autonomous vs teleoperated, measured?
6. With the founder and CTO gone, what is engineer retention on the hand/Carbon teams, and what's the lock-in/retention plan?
7. Are you positioning for an independent scale-up, a strategic acquisition, or an IP/licensing outcome — candidly?
8. What is the terms overhang from the Jan-2025 convertible (discount, preference) on the next round's common holders?
9. What is the Phoenix BOM today, and what unit volume reaches the ~$10–15/hr labor-payback threshold?
10. Why hydraulic when the field is standardizing on electric actuation — what's the durable advantage at scale and serviceability?
11. Is the patent estate encumbered (pledged against debt, licensed exclusively, tied to any investor)?
12. What does the Microsoft relationship contractually commit to — revenue, units, or just GTM optics?
13. How do you out-iterate Figure/Apptronik/Tesla on the data flywheel with a fraction of their deployed fleet?
14. What is board composition and control now, and which investors hold blocking rights post-purge?
15. What is the 3-year survival plan if no priced up-round materializes in the next two financing windows?

---