Join the movement and participate on Uniswap.
Participate
Nils is back in the Hong Kong lab. The demo store rebuild is mostly done: bigger, more realistic, and about to start taking data. Robin's Zappar scanner IDs a whole bay of markers in one frame, and that clip put a stuck digital-twin conversation back in motion.
We're teaching phones and robots to read custom, robot-friendly shelf markers: small, and legible at a distance. Hit record, and the system IDs every marker in the frame at once.
When we started rolling out Cactus, we scanned one price tag at a time. Robots don't move as quickly or as fluidly as people, so that pace does not work for them. Doing more with the same image feed is how a robot builds the twin without crawling the aisle SKU by SKU.
"This capability to do more with the same image feed is a really big deal for having the robots do this kind of data collection in the store."
That clip unstuck a lead. The retailer has 500 locations in the core brand and another 1,600 in the group. They liked the software. They did not like the cost of keeping a digital twin current if you scan products one by one. We sent the latest scanner video. We're back on track.
Shout-out to Robin, and to Zappar for the marker work.
We're talking to at least three large global ESL providers about showing these markers natively on electronic price tags. They already sit with a lot of retailers, and they're under pressure to improve what the tag can do.
Don't expect those deals to move like a Silicon Valley startup. ESL companies don't.
Live question from X: can existing ESLs show these markers, or do stores need new hardware? Largely a software update. E-ink price tags can display almost anything. Most are one color; four-color tags are getting more common. How long integration takes is case by case — some chains run the vendor's software, some run their own.
Nils's least favorite example: a five-figure store count, one of the biggest chains in the world, and the ESL backend is remote desktop into a single PC running Excel.
"I still get trauma thinking from that."
Where the main robotics office used to be is now aisle. New kit is in; product is still going on some of it:
Bigger floor, more realistic scenarios, data collection in this facility. We'll use it to train, and to show the product.
The wheeled humanoid we've been incubating in the lab raised money — enough to ramp production and run larger pilots with us. We're not saying how much.
At internal demo day they showed a tele-op stack: drove over from the other side of the building, opened a door on the way, entered the demo store, and helped move a box.
They'll use this environment to train product fronting. We called this sequence more than a year ago: perception and answering questions first, then fronting and order picking. Those two keep a robot busy all day more easily than restocking, and they're an easier sell to retailers. The new store is where we start collecting that data.
Hopefully we can show the robot soon. There may be a path to do some of that data collection here in the lab.
The gas-leak monitoring project we mentioned recently is moving. We signed a first contract with the partner. First gas-leak detection robot delivers hopefully within a month.
They already run this job across a lot of retail locations. A robot is better value for their customer, better margin for them, and more nodes on the network for us.
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
