THE OPPORTUNITY
A $1.5M pre-seed to take SwUM from a Gazebo prototype to a field pilot.
Three engineers, 24 months of runway and a field pilot in about 7 months. The rails SwUM runs on are already live, the specification is checked, with 33 of 33 numerical checks passing, and an early prototype runs in Gazebo simulations.
$1.5M
PRE-SEED SAFE
24
MONTHS OF RUNWAY
$62K
AVERAGE MONTHLY BURN
100%
FOUNDER OWNED
WHY NOW
Three things changed at once.
Compute moved onto the robot, the link stopped being trusted, and buyers set a deadline. The proof layer SwUM needs did not exist a year ago. SUM Chain, SNIP and OmniNode went live in the last twelve months.
$47.4B
RAISED BY PHYSICAL-AI STARTUPS IN THE FIRST HALF OF 2026
~10,000
DRONES LOST A MONTH IN UKRAINE, MOSTLY TO JAMMING
6 GB
ALL SWUM NEEDS ON AN 8 GB JETSON ORIN NANO
$47.4B · CRUNCHBASE NEWS, AUGUST 2026
~10,000 · RUSI ESTIMATE VIA IEEE SPECTRUM, APRIL 2025 →
ONE
Compute moved onto the robot.
The Jetson Orin Nano is held back by its compute limits. SwUM fits inside its 8 GB and lets a fleet host several small agents across the robots instead of one large model in the cloud.
TWO
The link stopped being trusted.
Radios get jammed and Wi-Fi floods at fleet scale. A fleet that needs the link to think stops when the link goes. SwUM needs one hop to a neighbor, and nothing else.
THREE
Buyers set a deadline.
Every US squad fields expendable drones by the end of 2026, and VDA 5050, the factory standard, still has no robot-to-robot coordination. The people who run fleets are asking for this now.
THE MARKET
Priced per robot, a market of its own by 2030.
Coordination is priced per robot per month, plus a paid pilot up front. Foxglove and Formant already bill fleets per device, so buyers know the shape of the deal.
Interact Analysis expects 4.2 million mobile robots installed by 2030. Our assumption is that about a million of them run in multi-robot fleets, which makes the coordination layer alone a market of $240 to 600 million a year.
WHY WE WIN
A better-funded team can’t copy this in a year.
The proofs already exist. Exactly one robot starts each job (Theorem 4). One shared view on any lossy radio (Theorem 1). The same answer on every chip, bit for bit (Proposition 7).
The rails already run. SUM Chain is on mainnet with more than ten million blocks. SNIP storage is live with proofs of retrievability. OmniNode attestations are live. RUMUS is open source.
TRACTION
A prototype in simulation, revenue zero today.
An early SwUM prototype runs in Gazebo simulations, and the spec's thirteen results are proved, with 33 of 33 numerical checks passing. Nothing is signed and revenue is zero. With funding: three real robots in about 6 months and a field pilot of 5 to 10 robots in about 7. The product-market fit test is two of three pilots converting to paid contracts within 90 days.
USE OF FUNDS
Half of it goes into engineering.
WHERE THE MONEY GOES
