From Security Jamming Our Robot to Global Expansion

Terri vs. the RF Jammer in Malaysia

Nils isback in Hong Kong after the ASEAN AI Summit in Kuala Lumpur, where Terri, our Unitree G1 humanoid intern, went on stage with the Prime Minister of Malaysia.

Security had a radio frequency jammer running to block hostile drones. Unfortunately, it also blocked our control link to Terri. Result: he just wouldn’t let go of the phone he was supposed to hand over.

The clip made Malaysian national TV, everyone laughed, and we’ve already been invited back for another event. It also sparked a useful point: for robots to really work in the world, they need:

• Reliable connectivity
• Maps and perception of every venue
• Infrastructure that’s actually built for embodied AI

That’s exactly what we’re building.

Our Role: Make the Physical World Accessible to AI

For anyone new joining from Malaysia, Base, or elsewhere: Auki’s mission is to make the physical world accessible to AI.

We work on collaborative machine perception, mapping, and positioning, so devices like robots, glasses and phones can:

• Understand their environment
• Share maps
• Know where they are indoors (no GPS)

We don’t build foundation models. Instead, we’re a bridge to the physical world for every model. You can use Auki + OpenAI, Auki + Claude, etc. We give them spatial context and embodiment.

Mentra Glasses + Auki: A Turnkey Vision Pipeline

A big focus this week: Mentra’s programmable smart glasses and how they plug into the Auki network.

Glasses provide live visual context, but devs need somewhere to send and process that video. We’re building that pipeline:

1. Mentra glasses send images/video to domain servers (data nodes) on the Auki network.
2. From there, data can go to:
   • Reconstruction nodes (for 3D scenes)
   • Vision nodes (for computer vision tasks)

“You don’t want to do the compute on the camera. You want the compute to happen somewhere else, and Auki is providing turnkey infrastructure for that.”

Our team is working with Mentra both in Hong Kong and in San Francisco to make this pipeline open source and reusable. Expect:

• Grants for builders
• Early projects using Mentra + Auki
• Demo videos of new physical AI copilots

Hybrid Robotics: Shipping “Half a Robot” via Glasses

Most robotics companies chase locomotion and manipulation first (walking and grabbing). We’re doubling down on another strategy: “You can ship half of a robot in a pair of glasses and let the human be the arms and legs.”

We focus on:

• Perception
• Mapping
• Positioning
• Applications

Glasses + the Auki network = AI copilots for physical work, right now, without waiting for humanoids to fully mature.

Our own retail co-pilot, Cactus, is one example. More copilots from community builders are coming soon.

Decentralized “Nervous System,” Not a Data Hoover

A good question from the community: if we’re decentralized and respect data autonomy, how do we compete with giants like OpenAI?

Key points:

• We don’t train foundation models.
• We provide context and coordination, not another LLM.
• Venue data can stay on their hardware, under their control.
• We monetize the protocol that lets machines exchange data and compute, and we “tax” those transactions.

For non-sensitive workloads (like consumer 3D memory capture), reconstruction jobs can be sent to community GPUs. For sensitive commercial work, customers can limit processing to trusted machines or partial views (e.g., one shelf, not a whole store).

Reconstruction Nodes & Community GPUs

We’re preparing to launch reconstruction nodes and already have an early partner: a consumer app for capturing 3D memories (“Instagram for spatial memories”).

Each job:

• Runs on a single consumer GPU
• Finishes in under ~15 minutes
• Produces public content

We plan to:

• Route these jobs through community reconstruction nodes
• Subsidize early workloads from the foundation so operators can earn meaningful tokens
• Offer bounties to extend node functionality via open source contributions

Short, self-contained jobs make this a great first fit for community GPUs.

US Trip, Basecamp & San Francisco Office

Travel change: instead of Europe, I’m heading to the US.

• Basecamp (Base ecosystem meetup):
  I was personally invited and will use it to make sure the Base community knows we’re the robotics / physical AI play on Base.
• September 9 – LA retail event:
  I’ll be in Los Angeles to pitch Costco and other major retailers on how our spatial AI can transform their physical operations.
• New SF office with Mentra:
  We’re opening a shared space in Hayes Valley, San Francisco, with Mentra. Together we’ll have a joint showroom where people can try AI copilots for physical work in person.

Our main office remains in Hong Kong; SF becomes our US hub.

Denmark Enterprise Deal & a Global Hotel Chain

Commercially:

• Our pilot with a Nordic retailer (~2,000 stores total) is going well.
• We’re close to signing a deal for 100–120 stores across two of their brands, where our value prop and margins are clearest.
• If results are strong, we’ll expand to their largest brand afterward.

We also have an upcoming meeting with a C-level exec at one of the world’s biggest hotel chains to explore how physical AI and copilots can help them beyond the retail context.

‍

Watch the whole update on X.

アウキ・ラボについて

Aukiはポーズメッシュという地球上、そしてその先の1000億の人々、デバイス、AIのための分散型機械認識ネットワークを構築しています。ポーズメッシュは、機械やAIが物理的世界を理解するために使用可能な、外部的かつ協調的な空間感覚です。

私たちの使命は、人々の相互認知能力、つまり私たちが互いに、そしてAIとともに考え、経験し、問題を解決する能力を向上させることです。人間の能力を拡大させる最も良い方法は、他者と協力することです。私たちは、意識を拡張するテクノロジーを構築し、コミュニケーションの摩擦を減らし、心の橋渡しをします。

ポーズメッシュについて

ポーズメッシュは、分散型で、ブロックチェーンベースの空間コンピューティングネットワークを動かすオープンソースのプロトコルです。

ポーズメッシュは、空間コンピューティングが協調的でプライバシーを保護する未来をもたらすよう設計されています。いかなる組織の監視能力も制限し、空間のプライベートな地図の自己所有権を奨励します。

分散化はまた、特に低レイテンシが重要な共同ARセッションにおいて、競争優位性を有します。ポスメッシュは分散化運動の次のステップであり、成長するテック大手のパワーに対抗するものです。

アウキ・ラボはポスメッシュにより、ポーズメッシュのソフトウェア・インフラの開発を託されました。