Pinned Tweet
What we shipped this week.
- Lamp: a new chapter
- 50% off flash deals at: autonomous.ai
- Build with Gradium and love Lamp? We have some gifts.
Full logs below. Anything on your wishlist yet?
4/ Leave your Computer builds to us.
Another Computer 2 landed in California this week. Two RTX PRO 6000 Blackwell, 192 GB of VRAM.
More people want the model running where the work already is. The whole build is open source.
nitter.gq/dee_hw/status/21230620…
1
6
120
5/ Next on the building list.
Lamp was the first of its kind to come alive. This one takes us further.
A humanoid foot. For... what? We'll keep that to ourselves for now. 🤫
nitter.gq/dee_hw/status/21234214…
The Foot Problem
There are all kinds of interesting engineering problems at @autonomous_labs. One I enjoy working on most is designing a humanoid's foot.
A robot hits the ground thousands of times a day. Sometimes it's a gentle step, sometimes it's landing a backflip.
So we started from first principles: what loads does a foot actually have to carry?
We listed the worst cases and modelled each one:
- a backflip landing: 1,233 N
- standing on the tip of the toe: 913 N
- one toe corner taking the load: 489 N
- rolling onto the inner edge: 1,000 N
- a hard heel strike: 1,511 N
Then we designed the foot around those loads with four components:
- a 3 mm carbon-fibre plate that is stiff, carries load, and springs where it lands
- a 7075 aluminium lug block that ties the plate to the ankle
- six titanium screw cups
- a 4 mm rubber sole
The plate bends less than 2 mm under a backflip landing. Its peak stress is under a third of what the carbon is allowed to take. The whole foot weighs 117 g, 12% lighter than the aluminium frame it replaces.
Its ranges are amazing:
- toes up 45°
- toes down 30°
- tilting 25° each way
This is a super fun problem. If you're working on humanoid's foot, would love to exchange notes.
6
112
How we design a humanoid's foot ↓
The Foot Problem
There are all kinds of interesting engineering problems at @autonomous_labs. One I enjoy working on most is designing a humanoid's foot.
A robot hits the ground thousands of times a day. Sometimes it's a gentle step, sometimes it's landing a backflip.
So we started from first principles: what loads does a foot actually have to carry?
We listed the worst cases and modelled each one:
- a backflip landing: 1,233 N
- standing on the tip of the toe: 913 N
- one toe corner taking the load: 489 N
- rolling onto the inner edge: 1,000 N
- a hard heel strike: 1,511 N
Then we designed the foot around those loads with four components:
- a 3 mm carbon-fibre plate that is stiff, carries load, and springs where it lands
- a 7075 aluminium lug block that ties the plate to the ankle
- six titanium screw cups
- a 4 mm rubber sole
The plate bends less than 2 mm under a backflip landing. Its peak stress is under a third of what the carbon is allowed to take. The whole foot weighs 117 g, 12% lighter than the aluminium frame it replaces.
Its ranges are amazing:
- toes up 45°
- toes down 30°
- tilting 25° each way
This is a super fun problem. If you're working on humanoid's foot, would love to exchange notes.
2
528
If you build with @GradiumAI, the real-time voice AI for agents and devices, Black Friday came early.
20% off Lamp, through Nov 30. Your code is waiting in the Gradium community.
autonomous.ai/lamp
Feeling bored at night?
“Little assistant, play the drama!”
Yep. Not just an AI assistant, Intern can now also play radio dramas.
Mystery, sci-fi or romance tonight? 🎧
Lamp Skill Store 👇
Physical Skills
Lamp is a physical AI agent. And Lamp's got physical skills. Dance. Security Guard. Drink Water Reminder. And many more.
autonomous.ai/lamp
This week, we're focused on shipping the Skill Store (aka the App Store for Lamp). So you can easily install a new physical skill for Lamp.
Another design goal is to make it super easy for you to prompt a new physical skill. As easy as prompting the agent skills you already know.
You can keep the skill for yourself or you can publish it for others to use.
4
704
Another Autonomous Computer 2 just landed on our customer’s desk in California.
2× NVIDIA RTX PRO 6000 Blackwell. 192 GB of total VRAM.
Planning to run AI locally? Tell us what you’re building and we’ll help you spec your rig.
Lamp will be there. See you all in Shenzhen!
你好 Shenzhen, OpenClaw’s @clawcon is coming in December openclaw.ai/blog/openclaw-sh… @themu_xyz @M5Stack @TencentGlobal @seeedstudio @pamir_ai @vllm_project @autonomous_labs @sspai_com
4
715
Autonomous retweeted
Replying to @codyplof
@autonomous_labs standing desk, Edifiers, Asus ROG monitor, and a mountain view is all you need.
Don’t be like me. Get a real camera.
I never imagined I'd spend a beautiful Sunday teaching a robot about the "kinetic chain".
Pretty fast learner. Within a couple of hours, it was hitting 6x harder with a trained policy. It learned to pivot its rear foot and turn its hips to generate power.
If you don't box, the kinetic chain is how a punch gets its power. It starts at the ground and travels up through the legs, hips and shoulder, rather than coming from the arm alone.
Not bad, buddy!
What we shipped this week.
- Black Friday is on
- Just launched: Autonomous Radio, a radio full of stories
- Lamp and Computer reached new homes
Full logs below. Anything on your holiday wishlist?
5/ Time to be your own datacenter.
Another customer emailed us a photo of their new Computer 2. Two RTX PRO 6000 Blackwell GPUs, right on their desk.
nitter.gq/dee_hw/status/21201593…
A datacenter in every home.
A customer emailed us this photo today. Their new Autonomous Computer, built with 2x RTX PRO 6000 Blackwell GPUs, right on their desk.
Just super happy to see our builds out in the world helping people own their compute!
autonomous.ai/computer
1
5
162
6/ Spaceship for GPUs?
We launched Spaceship, a quiet room for you and your agents at the office.
Should we build another one for your GPUs?
It has to stay cool, quiet and shared across the team. We're thinking it through.
nitter.gq/dee_hw/status/21205353…
A datacenter in every office.
We've built a lot of personal datacenters for people running local AI at home. I've been thinking a lot about bringing that to the office.
We keep hearing the same things from our customers and partners. They want to:
1. Own their GPUs
2. Stop paying per token
3. Run without rate limits
4. Keep their data in the office
5. Share compute across their entire team
The biggest challenges are cooling and noise. It has to be something people can actually work next to.
I think we can solve those problems.
4
137
















