Join the movement and participate on Uniswap.
Participate
Nils dialed in from Stockholm after a headquarters visit with one of our key customers, mapping robot rollouts for early next year. Cactus — the retail copilot already running in stores — is about to get robots that pull context from the floor on their own. We're two to three months from loading those machines into shipping containers for the first Scandinavian pilot sites.
The hard problem is no longer teaching a robot a trick on camera. It's getting the same robot to do the job every day without a developer in the aisle.
Once a robot hits a site, it has to know which store it is, which customer owns it, and which servers to talk to. Our sites already run digital twins for the copilot. That is the advantage.
Here's a walkthrough of the onboarding flow we keep tightening:
Cool demos are easy to film. Cutting setup time, maintenance load, and hands-on touch from the robot developer is the product work that actually ships.
This week the team showed the first end-to-end over-the-air update path. When something in the stack needs to change, we can push code the way a phone takes an OS update — remote, without a site visit for every fix.
In parallel, Jason's fleet management work is landing inside Cactus:
We've also hired two people for this surface — one starts next week, one about two weeks after — specifically to get ready for robots that live off-site.
Phil put it plainly from the robotics side: the skill itself looks less like the bottleneck every month. Interfacing, monitoring, updating, and deploying look more like it.
Phil spent the week in product legislation. CE marking and the EMC story are the simple end — keep your Wi-Fi from wrecking someone else's radio. Robots sit under the Machinery Directive, which gets replaced in January with a new standard that starts looking at embodied AI and automation. Lots of open questions.
Most of the center-stage humanoid circuit still barely talks about the rules they need to clear before a real ship. That gap will get loud as people try to move metal out the door.
Nils also flagged a conversation with a senior FDE at a well-known American robotics company — one previously tied up in FCC robot bans, now working with another formerly American name on ISO-style standards for robots that can fall over. Charitable read: clearing purchase blockers. Less charitable: regulatory capture and advance notice on the rules competitors will have to meet.
Demos convinced a lot of people. The remaining laundry list — contracts, insurance, compliance, remote maintenance — is what decides who actually deploys.
Sam Altman went around again on wrong AI adoption timelines. The pattern maps cleanly onto robotics: the capability to program useful work is here at much lower cost than a few years ago. Diffusion still takes months even when the ROI case is obvious. Tech companies adopt fast. Most of the economy does not.
Retail is a sharp version of that. Geopolitics, supply-chain shock, and inflation made a lot of chains defensive — keep yesterday working before funding tomorrow. One retailer we expected to have closed by now looks like a couple more months. They went from expecting a strong year to a rough quarter. Compressed margins are an opening for a harder future-proofing pitch; they are also why a lot of 2026 moved slower than the lab calendar wanted.
Tracy's internal rollout manager made the same point from the inside: the real blocker list is longer than gut feel. Learning that list by trying to ship is the valuable part.
Anthropic's MHS work is a thin cut of what we're after with the new Auki SDK: machines that AI can actually drive. Lab gear under model control is a real result. What we're building with Intercognitive — machines that are not only controllable but discoverable and composable in collaborative workflows — is the thicker version of that stack.
If the Discord wants it, we'll schedule a ~one-hour technical deep dive in about a month on design philosophy, progress, and what's next. Flag us there.
Nils flies back to Hong Kong in two days to rebuild the lab demo environment into a more realistic supermarket: glass doors that open and close, reflections, mixed shelf types, freezers — a harder perception and navigation surface than the rough retail mock we had. Local academic work on next-generation collaborative navigation is also in motion; more when we can talk about it.
We run these on X every week. The useful half of the conversation usually moves to Discord for questions, early footage, and gossip. Join us at discord.gg/auki.
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
