For investors
The robots
are funded.
Their memory
is not.
The Physical AI stack has transport, training, space, and safety. It has no layer for what a machine learns after it ships — who owns it, how it transfers. benned builds that fifth layer.
Request investor deckThe wave is funded
Others are paying for the robot wave. Figure is valued at $39B. Roughly 16,000 humanoid robots were deployed in 2025. Goldman Sachs projects a $38B humanoid market by 2035, and Bank of America expects unit costs below $17,000 by 2030. Agility rents Digit at $10-12 per hour against a $30 human. Hardware is commoditizing on someone else’s balance sheet. We do not build robots. We build what every robot will need.
The missing layer
The Physical AI protocol stack is forming fast: ROS 2 and MQTT for transport, VLA models and LeRobot for training, OpenUSD for spatial context, Halos for safety. Four layers. None of them covers deployed knowledge — what a robot learns after it ships, who owns it, and how it transfers across vendors and machine generations. Every layer of the internet stack minted an infrastructure company. This one has no owner yet.
Kin — the layer
Kin is a persistent, owned, transferable knowledge entity per person, site, or fleet. It lives in sovereign cloud or on-premise; the customer holds the keys. Machines connect and inherit what the entity knows; what they learn flows back. Kin builds on the existing standards rather than competing with them. Platform economics: per-entity, compounding — the more a Kin knows, the more it is worth, and the harder it is to leave.
Lore — the wedge
Lore is capture software that sells today, before the robot wave peaks. A skilled worker records while working — video and voice, with a phone or an existing bodycam. Lore structures it into searchable task knowledge inside the customer’s Kin. The buyer’s problem is immediate: the caretaker with 25 years of building knowledge retires and nothing is written down. Knowledge retention pays for the product now; every capture builds the data moat for when machines arrive.
Why now
The EU Machinery Regulation enters enforcement on January 20, 2027 and forces documentation of machine behavior. ISO/WD 26264-1, the draft standard for humanoid datasets, shows standardization is coming — but it covers dataset architecture only, not runtime knowledge or ownership. And every robot sold by someone else makes the knowledge layer more valuable, at zero capital cost to us.
The moat
Neutrality first: a knowledge layer owned by a robot manufacturer is exactly the lock-in customers are trying to avoid. benned sells no robots, so we can be the layer every vendor connects to. Second: accumulated field knowledge per customer. A Kin that holds ten years of a site’s operational knowledge is not something a competitor can replicate with a bigger model. The data is the customer’s — and the customer stays where the data lives.
One AI.
Every machine.
Kin in every device and robot. Lore keeps what your people know. No setup. No apps. No starting over.