LUCKEY · Seed Memorandum · April 2026
LUCKEY.
Mercor for Physical AI — and the consumer brand that grows out of it.
Shawn Gong · CEO · shawn@luckey.to
Forbes 30U30 · Red Dot BoB · Nums 1M+ shipped
Founders Inc '26 cohort · just selected
Forbes 30U30 · Red Dot BoB · Nums 1M+ shipped
Founders Inc '26 cohort · just selected
$5M SAFE · $20–30M Post
TSVC preferred lead
TSVC preferred lead
01 / 12
02 · Problem · Why Now
Embodied intelligence
isn't on the internet.
LLMs scraped the web. Robots can't scrape what was never written down.
"We will need orders of magnitude more data than we have today."
— Karol Hausman · CEO, Physical Intelligence ($400M raised)
01 Buy side $4B+ · PI · Skild · Genesis
02 Scaling proven · NVIDIA EgoScale R²=0.9983
03 Teleop $75–100/hr · 50% waste
04 Wearable BOM $30 · was $200+ in 2024
02 / 12
03 · Solution
Purpose-built for
egocentric data.
Action cams record what humans share. $70 LUCKEY records what robots learn from.
GoPro · $400 · 50% data wasted
Build.ai · factory only
Build.ai · medical look
LUCKEY · $70 · 95% data utility · purpose-built
03 / 12
04 · Customers · Traction
$100K first-wave LOIs.
Everything HA's $6.5M was built on — plus more.
01 · Direct · LOI signed
Orbifold + Linkwise
Orbifold · 200 × $70 = $14K
Linkwise · 1000 × $70 HW + 500hr × $30-40 data = $85K
Linkwise · 1000 × $70 HW + 500hr × $30-40 data = $85K
02 · Their downstream → ours
Linkwise → Tesla · Figure · 1X
Orbifold → Sunday · World Labs
Orbifold → Sunday · World Labs
Direct, Mar '26 sprint to core teams / data procurement: Genesis AI · Luma · Runway · Dyna · Chestnut (CTO advisor) · Dexmate (via Xia Chun) · 智元 / AgiBot (CN Tier-1)
03 · Workforce channel
Wiz.AI — 18 countries, 300+ enterprise
Multi-region BPO labor pool US-only competitors structurally can't access. Combined with LUCKEY's $70 unit cost, data acquisition drops ~10× vs industry norm.
"Workers forget to hit record. 30–50% of current data is unusable."
— Enoch Zhu, CEO Orbifold · told us directly
04 / 12
05 · Moat
What vibe coding
can't replace.
Hardware. Data. Cost. Wearable. Only earned via B2B → B2C, in that order.
01 · Hardware
Orbifold CTO co-spec'd. Build.ai-validated supply. Nums 1M+ shipped.
Spec reverse-engineered from Tesla / Figure / Sunday demand. 7 form factors on one stack. 85 patent claims filed (PPA 63/999,137).
02 · Data
B2B Tier-1 spec + B2C multi-region diversity.
Wiz.AI 18-country workforce — hospitals, banks, BPO, call centers. Cofounder-level trust (not vanilla BD) makes this structurally non-replicable.
03 · Cost
Industry pays wearers $75–100/hr. We charge $70, then collect.
~10× acquisition cost drop vs industry norm. Negative-cost data — structurally inverted.
04 · Wearable
Always-on. Worn voluntarily. Agent's eye + user surface.
Phones miss the action (in pocket). Glasses see forward only. Same hardware → iPhone-class terminal for AI services on top.
Negative-cost data Industry: $400 device + $75-100/hr labor → LUCKEY P1 (B2B): customer pays $70 + data revenue → P2 (B2C): customer-paid + passive capture → NEGATIVE cost.
05 / 12
06 · Form Factors
One stack.
Seven form factors.
Bay Area's CTOs spec'd this for robots. Consumers want the same architecture.
Now
Visor / Clip
Q4 '26
Ear-Clip
Q4 '26
Sport Shades
2027
Ski Goggles
2027
Dive Mask
2027
Golf Cap
2027
Helmet
Step 1 · Now B2B Visor/Clip
Step 2 · Q4 '26 Consumer forms
Step 3 · 2027+ Largest diverse dataset
Step 4 · 2028+ AI services on same hardware
06 / 12
07 · Robot's YouTube
Robot's YouTube.
10 minutes for human YouTube. 90 minutes for the robots that learn from us.
01 · Creator
Hyejee Bae
Harvard '20 · 250K subs · Creator Advisor
"btw how can I get one of your devices?"
LinkedIn DM · April 2026
Solves → Multi-cam · 3rd-person POV · Solo production · Hands-free · Waste footage → income
02 · Safety · Awareness
Side-rear blind spot for cyclists, runners, walkers.
Garmin Varia rear-view radar sells $1000+ as a single-purpose device. We do it integrated — multi-sport, multi-form factor.
03 · Data flywheel
More wear → rarer data → smarter robots.
A person falling while interacting with a robot — scarce in factories, abundant in consumer life.
+ Health behavior
Diet log (chew→cam) · CGM correlation · Sleep patterns · Longevity protocols.
In conversation with Stanford longevity lab.
Tesla: drivers → negative-cost data at scale → best FSD. LUCKEY: wearers → negative-cost data at scale → physical-world intelligence.
Hardware and data capture are the means — intelligence is the goal.
Hardware and data capture are the means — intelligence is the goal.
07 / 12
08 · Strategic Optionality
Generational standalone.
Plus three acquirers.
Same hardware. B2B data revenue funds the standalone AI Agent platform — or any of three industries that already need this acquires.
Standalone endgame
LUCKEY = AI Agent hardware platform.
B2B data revenue funds B2C AI Agent rollout. Same hardware. iPhone → Apple model. Generational outcome.
Acquirer 1 · Data infra
Scale / Mercor expanding into Physical AI.
They label. We supply.
Acquirer 2 · Robotics
Tesla / Figure / PI vertical integration.
They are the buyer. Easier to acquire than build.
Acquirer 3 · Consumer HW
GoPro / Apple / Meta consumer SKU.
Same architecture, post-traction.
"I build the company as a product. I think about who will buy us in the future."
— Shawn Gong
08 / 12
09 · Team
Bay Area + Shenzhen
+ six named advisors.
Core · Full-time
Shawn Gong · CEO
Forbes 30U30 · Red Dot BoB (youngest, 19) · 34 patents · Tsinghua · Nums 1M+ shipped
Mingmin (Echo) She · Co-founder
Ex-Google 7y (HW & SW, Mountain View) · UCLA Anderson MBA · Arrives US May with first batch
In Conversation · Joining 3–6 mo
Ray
CMU HCI · Bay Area · wearable UXKeren Ni
Tsinghua / ex-Huawei Watch GT designerKazuma Hakushi
Berkeley · ex-Mercor · MLHyejee Bae
Harvard · 250K subs · Creator AdvisorSix Named Advisors
Dr. Michael Barry (Stanford d.school) · Julian Quevedo (Stanford CS, ex-World Labs) · Lianfeng Xiao (30y Sony optical) · Yilin Zhu (Apple Vision Pro) · Yongtao Xie (NY Bar, data privacy) · Yuhang Shi (Oxford DPhil, 3D)
09 / 12
10 · Pricing Anchor
Everything HA has.
Plus three more.
HA's customer face. Build.ai's supply ecosystem. Plus hardware shipping, a Nums-shipping founder, and a consumer roadmap.
| LUCKEYFounders Inc '26 · Raising $5M / $20–30M post | Human ArchiveYC W26 · Raised ~$6.5M | LinkwiseYC · Raised $1.1M | Build.ai / Xiaomi-CNRaised ~$15M @ $150M post | |
|---|---|---|---|---|
| Named customers | Orbifold + Linkwise LOI | 1 — Linkwise | Tesla · Figure · 1X | Factory or consumer |
| Hardware shipping | Ships May · Red Dot BoB | Vague rig | Data ops only | Factory / app-first |
| Founder track record | Nums 1M+ · 34 patents | First-time | Factory-training pivot | First-time / Xiaomi alums |
| Consumer roadmap | Agent entry · 7+ forms · creator/safety/health | None — pure B2B | None — pure B2B | App-only |
| Market validation | Capital has validated the category — $22M+ raised across HA / Linkwise / Build.ai. LUCKEY brings what they don't: shipping hardware, Nums 1M+ founder, dual B2B + B2C reach. | |||
Buy-side TAM Generalist $140M · Genesis $105M · PI $400M · Skild $4B · Tesla Optimus · NVIDIA EgoScale
10 / 12
11 · Pressure Tests
Real. Durable.
Defensible.
Real
End-to-end loop unproven?
Orbifold's loop runs today on $400 GoPros. We substitute, we don't reinvent. Model improvement is downstream (Tesla / Figure) on a 6–12mo lag.
Durable
10-year data decay?
Strategic optionality (slide 8). 3 industries already need this. Consumer pivot lives on the same hardware.
Defensible
Hardware commoditization?
Hardware is the entry ticket — not the moat. 4 pillars (slide 5) require 1–2 years of relationships to copy. Plus 85 patent claims (PPA 63/999,137 · 16 subsystems) + LUCKEY® US-registered.
Software is no longer the moat. Hardware + proprietary wearable data + consumer trust — LUCKEY has all three. Only accessible via B2B → B2C, in that order.
11 / 12
12 · The Ask
$5M SAFE.
$20–30M post.
TSVC preferred lead.
Use of Funds
35%
Data pipeline & AI infrastructure
Sync · anonymization · pose · curation · model training
25%
Team expansion
Scale to 8–10 FTE
20%
Hardware iteration & manufacturing
Mature supply chain · batch production
20%
GTM & customer expansion
Bay Area buy-side + SEA via Wiz.AI
6-Month Milestones
M0
Orbifold PO signed · first payment
M1
Echo + first hardware batch arrive US · Linkwise 500hr Philippines BPO data ops live (first data revenue)
M2
2nd direct LOI converted (US Tier-1 · Chestnut / Dyna / Genesis AI shortlist)
M3
Linkwise 1000-unit production · custom hardware spec locked
M4
1st CN Tier-1 LOI (智元 / AgiBot) · multi-account data trials (Sunday / Luma / Runway)
M6
First B2B vertical pilot · first creator content · downstream model-improvement signal (via Linkwise → Tesla / Figure)
Why TSVC Eugene's 2-yr founder relationship · Kai's Google/AI network · Global Shutter portfolio · Dexmate · co-sponsored by Shen Kai + Xia Chun
12 / 12