Join the movement and participate on Uniswap.
Participate
We ran an hour late from Hong Kong — still clearing more than 100 people out of Robots & Beers. The open house was good. The lab milestone mattered more.
For the first time end to end, Cactus took control of a robot, chased a product problem, and closed the loop without a human on the sticks.
Phil’s team has been wiring a robot store manager into the same AI brain already scaling with major retailers. This week the full chain worked.
In the demo, a colleague talks to Cactus at a shelf. One SKU is soft: Lay’s sour cream and onion. He asks the store AI to investigate. Cactus commandeers the robot. The machine drives to the bay, aims a wrist camera, and reports back.
Empty shelf. Inventory said the product was still on the floor. It wasn’t — short shipment, shrink, or both. Inspect-product completes. Cactus’s call: restock this slot immediately. Price and eye-level position are fine; availability is the failure.
“I could be a store manager in my vacation house in Spain… and the robot drives there on its own.”
A human asked the question this time. The real product is the opposite: the AI asks on its own — sales anomalies, empty facings, planogram drift — and investigates overnight while staff sleep. Autonomous investigation is the next demo, hopefully within a week or two.
A supermarket runs tens of thousands of products. No human solves that optimization at full fidelity. CCTV, phones, glasses, and robots feed one store brain that allocates shelf space and staff time better than any single manager can.
We do not need arms restocking for the unit economics to work.
Customers are offering on the order of $40,000 per robot per year — roughly $100/day — for eyes and ears on the store AI. Perception only. No manipulation required at that price.
External context, not our result: a Chinese retailer building similar software in-house lifted sales 258% without robots. Robots feeding the capture layer should compound that class of gain.
On deployment cost: the demo robot never mapped the store itself. It inherited the twin the AI already had from staff phones. For customers already on the phone copilot, robot deploy is under two minutes.
Demand behind next year’s numbers: a handful of customers already asking for 800–1,400 robots each.
We said first units by end of July. We missed.
What slowed us: robot insurance, financing, pilot cost. Enough of that is sorted for November containers. Not everything is finished — enough is.
Second clip: our camera-based barcode reader. As far as we know, it is among the fastest camera barcode scanners shipping. The replay is not sped up — real capture rate, fast enough that drive-by reads on a moving robot look plausible.
We planned dedicated barcode hardware on the robots. This pipeline is good enough that we may drop those scanners. XR roots show up here: augmented reality and robotics share more DNA than the industry usually admits.
Twins still have to stay live while humans rearrange the floor daily. Maintenance paths: robots, phones, and Mentra glasses (Nils was wearing them on stream). Same brain, more sensors.
Auki is making the physical world accessible to AI by building the real world web: a way for robots and digital devices like smart glasses and phones to browse, navigate, and search physical locations.
70% of the world economy is still tied to physical locations and labor, so making the physical world accessible to AI represents a 3X increase in the TAM of AI in general. Auki's goal is to become the decentralized nervous system of AI in the physical world, providing collaborative spatial reasoning for the next 100bn devices on Earth and beyond.
X | Discord | LinkedIn | YouTube | Whitepaper | auki.com
