The Robotics Stack

Weekly podcast on what's actually shipping in robotics.
Hosted by @hundreddaysx and @leokharon

[The Robotics Stack @roboticsstack]

VHS, DVD, the internet, VR glasses. Somnia CEO Xara says sex was the first product-market fit for all of them.

Now it's 2026 and the company is building sex robots.

We asked about it on @roboticsstack

New episode 👇

https://t.co/1yMop3iJ5O

[The Robotics Stack @roboticsstack]

Billions are going into factories to build tens of thousands of humanoids.

Matt Trossen has sold robot hardware for 22 years and cannot name one customer who wants to buy one.

https://t.co/4xl11oEC4H

[The Robotics Stack @roboticsstack]

You want to train your humanoid fully in sim and deploy zero-shot.
Then your actuator behaves nonlinearly and the whole plan falls apart.

We sat down with the CPO behind Asimov, the open-source humanoid, on @roboticsstack

New episode 👇

https://t.co/5eCTtvMKff

[The Robotics Stack @roboticsstack]

An openai intern gave GPT-6 Astra a cheap 3d printable arm, a paintbrush and two cameras, then told it: paint the golden gate bridge.

Astra worked out the arm itself, then started correcting its own errors.

Six astra robot demos on @roboticsstack

https://t.co/WclzpKnInr

[The Robotics Stack @roboticsstack]

Numb your fingers and good luck trying to light a match. That's a robot hand today

Matt Wulff @makeitmatto, cofounder of @6thSensedev, on why he thinks touch is the missing modality, studio-recorded data is fishy, and UMI is not here to stay.

New episode
https://t.co/yhtMTqE3oZ

[The Robotics Stack @roboticsstack]

What happens to US robotics if Unitree gets banned?

Randall Briggs left MIT to build Steel Bot, an open-source, made-in-America humanoid.

We talk specs, modularity, hands, funding, and what it really takes to mass produce one.

New episode out now 👇

https://t.co/btUWH0SanK

[The Robotics Stack @roboticsstack]

XPENG's Iron has the most natural walk of any humanoid so far.
Human-like spine, bionic muscle structures, flexible skin, and 900M raised!

Also: teaching a robot from one demo, Exotec's warehouse robots, and China landing its first booster on legs.

https://t.co/HI1SpTIXvd

[The Robotics Stack @roboticsstack]

A chinese humanoid company is now selling an AI companion with silicone skin, 88 DoF, and an LLM that reads 20 emotions.

13,000 orders at launch.

We dive into the AI companion market on @roboticsstack, including the cyber brothel that opened in Berlin.

https://t.co/rb7Xy0S7Ka

[The Robotics Stack @roboticsstack]

A chinese humanoid company is now selling an AI girlfriend with silicone skin, 88 DoF, and an LLM that reads 20 emotions.

13,000 orders at launch.

We dive into the AI companion market on @roboticsstack, including the cyber brothel that opened in Berlin.
https://t.co/lwqu3fOfO3

[The Robotics Stack @roboticsstack]

The Asimov 1 by @asimovinc might be the first real open-source humanoid.

$20k for the kit, $499 to pre-order, 100+ hours to build it.
Employee-owned, no VC.

Also: @DynaRobotics's new model, @trossenrobotics, and thoughts on robotics margins.

https://t.co/5xjbzXipOn

[The Robotics Stack @roboticsstack]

Gemini Robotics 2 can unscrew a light bulb 92% of the time.
But screwing it back in? 36%.

We cover Google's new models, + the top 10 robotics companies by revenue, why plugging ChatGPT into a robot doesn't work, and robot fights.

New episode 👇
https://t.co/HNUiqaDrJH

[Léo @LeoKharon]

Every humanoid joint you've seen is three motors stacked on top of each other, each one carrying the weight of the last.

@roboticsstack sat down with Bryce Cianciotto @CianciottoBryce, who built the Orbit Actuator that does all three axes with the motors at the base, and the wires routed through the middle.

New episode 👇

[The Robotics Stack @roboticsstack]

Every humanoid joint you've seen is three motors stacked on top of each other, each one carrying the weight of the last.

We sat down with @CianciottoBryce, who built the Orbit Actuator that does all three axes with the motors at the base.

New episode 👇

https://t.co/F7goB7DIbp

[The Robotics Stack @roboticsstack]

https://t.co/4WnonA1Jsh

[The Robotics Stack @roboticsstack]

An Italian startup just dropped one of the most impressive humanoids we've seen.
We dig into Generative Bionics' GENE.01, Sunday Robotics folding laundry with ACT-2, and the rumor that Anthropic tried to acquire Physical Intelligence.

New episode of @roboticsstack 👇

[The Robotics Stack @roboticsstack]

The hardest thing to build in robotics isn't the brain, but the hand 🖐️

This week, we get into 1X's NEO and why dexterous manipulation is the real frontier.
Plus: surgery bots, autonomous construction, and how you even benchmark a robot.

New episode 🦾

https://t.co/VmehQ7WYEu

[The Robotics Stack @roboticsstack]

Interesting how embodiment and intelligence condition ond another

[Léo @LeoKharon]

Robotics essentials: EMBODIMENT

A robot's intelligence is not only stored in its head. Part of it sits in the shape of its body.

The oldest proof is the passive dynamic walker: a pair of legs with no motors, no sensors and no controller at all, which walks down a shallow slope on gravity alone. The gait is not computed. It falls out of the geometry. Build the body right and it does some of the thinking for you, for free. The Cornell biped built on that principle walks at roughly human energy cost. Honda's ASIMO, computing every step, used about an order of magnitude more.

The flip side is the trap the field is in now.
A policy learns the body it grew up in: its joint count, its reach, its camera placement, its control rate. Move it to a different robot and it doesn't degrade, it dies.
Tested on robots they weren't trained on, every open vision-language-action model collapses to zero success. Not worse. Zero.

The escape plan is cross-embodiment: pool data from many robot bodies and train one model on all of them. Open X-Embodiment did it with 22 robot types and a million demonstrations. It works, unevenly, and how unevenly is a post of its own.

Which is the real argument for building human-shaped robots, and it is not about the hardware.
It is that the human body is the one morphology with a training corpus already sitting on the internet. Nvidia pretrained on 20,854 hours of human head-camera video; going from 1,000 to 20,000 hours more than doubled what the robot could finish. On the same task, one hour of human hand video beat one hour of robot data.

[The Robotics Stack @roboticsstack]

@MatthewBerman Isn't Poolside supposed to be French?

[The Robotics Stack @roboticsstack]

What can an AI learn from rocket league?
We dig into MIRA, General Intuition's new world model trained on
game data.

Plus Flexion's long-horizon leap, @tnkrdotai the "Github for robots", and Boston Dynamics at the World Cup.

New episode 👇

https://t.co/BwrgOVsQVA

[The Robotics Stack @roboticsstack]

They literally named a robot the "T800", after The Terminator.
We dug into EngineAI's new humanoid, Anduril's Arsenal OS, and the Tesla lawsuit nobody's talking about.

New episode 👇

https://t.co/NsuqE7oUwI