Hello, real world We're making the physical world accessible to AI

Earth and beyond
Centralized compute built the last decade of AI. It won't run the robot decade. @vveitas and Pravar Joshi of @nunet_global tell @RichardRobinson how complexity science and @bengoertzel's BGI bet will become the decentralized infrastructure robots actually use. Enjoy.
1
10
1,341
Like the internet, in the coming years "low altitude economy" is going to go from a fringe concept to daily necessity
A flying skateboard, and it's actually shipping! 🛹 Personal eVTOLs so far have meant a seat, a cockpit and a pilot's licence. It is a platform you stand on. Four ducted propulsion units arranged in petal-like sections, eight rotors, and a self-developed flight controller called the NA80 running dual-redundant architecture with three inertial measurement units per unit. The control scheme is the part worth understanding. High-precision sensors track the combined centre of gravity of rider and machine, and adjust rotor thrust as you lean. Steering is a nudge on a control stick, takeoff is a button press. CoolFly says most people get it in about 10 minutes, roughly like learning a self-balancing scooter. What makes this more than a stunt is the paperwork around it: China's revised Civil Aviation Law took effect on 1 July giving the low-altitude economy legal standing, and in June an insurer underwrote the first policy written specifically for standing manned flying vehicles. The law and the insurance arrived before the vehicle did. 11 minutes of airtime and nowhere to sit. Would you step on? :) ~~ ♻️ Join the weekly robotics newsletter, and never miss any news → ziegler.substack.com
1
53
Intercognitive retweeted
Running with WendyOS
1
2
9
427
I spend my days doing this but for dexterous robot data
By popular demand, I present to you.. my magnum opus.
1
58
Intercognitive retweeted
Two years ago I tried to raise awareness of how far China's robotics industry had come, and was treated a like a lunatic at the time. augmented.vision/global-shif…
4
2
46
1,912
Intercognitive retweeted
This guy lost four rounds before to the clanker. He only won this round and posted it
We put this on the hardware!! So excited that I won this round!!
3
9
123
14,705
Intercognitive retweeted
The fastest way to learn robotics is to start building the robot.
Day 3 of Building a Humanoid Robot. Head + Hands + Body half way
2
10
69
5,397
『造船所で溶接や検査を行う四足歩行ロボット』 足元に電磁永久磁石を採用し、垂直な壁面や曲面状の鋼構造物上の移動が可能。 lnkd.in/gtdePAqF #quadrupedal #robot #PhysicalAI #welding #inspection #shipyard #magnet #EPM #DidenSpider #DidenRobotics
68
If you're in HK and interested in robotics!
luma.com/cxhb5ij1 Plenty of warning this time so no excuses... Drop me a message if you have something you'd like to present, whether it is a paper, a project or just a rant about the industry.
48
Intercognitive retweeted
Replying to @Blockworks
@Blockworks powers the data behind our IR reporting. We're building the rails for physical AI, and we hold the numbers behind that to the same reporting standard as the rest of the infrastructure we run.
Replying to @Blockworks
Just in the last week: - @circle used our data on its earnings call. - @WSJ featured our Hyperliquid data. - @KobeissiLetter shared our DEX data. - @GEODNET embedded our data in its IR page. The industry has matured, and crypto executives, developers, and investors demand accurate, standardized, and comparable data. With that, the demand for accurate, comparable crypto data keeps growing.
5
10
70
4,223
Intercognitive retweeted
The idea of a collaborative perception network was a crazy unlock.
We've built most advanced AI system for retailers in the world. Now with autonomous robots, AR navigation and an AI copilot that helps you run the store.
3
12
84
2,780
Intercognitive retweeted
More robots coming out of stealth
We built a new kind of robot for the underdogs. Today, robotics is stuck between ~$100k systems and low-cost robots not meant to survive deployment. We think builders deserve better. Our mission is to develop the best hardware platform for others to deploy on. Meet Feather, an affordable general-purpose robot designed for real work: • 1m reach • Human strength • 10-hour battery • Shipping for the past year + available today Founded by @whoishoa and Parsa Bakhtiari, Feather is coming out of stealth with $7.6m in pre-seed funding led by @GradientVC, with participation from @BuilderVC, @geometryvc, @SEEDInnov, and Virgo VC. We’re hiring across engineering, ops, and go-to-market. If you’re an absolute maverick, join our cause today – feather.dev
2
6
127
15,142
Intercognitive retweeted
We put this on the hardware!! So excited that I won this round!!
Will release something interesting in the coming months.
26
89
609
78,644
Proposing a new PGI benchmark: PABE (Peeling A Boiled Egg) If your humanoid can't do this without shredding the white, we do not have physical general intelligence Full disclosure: I need this more than anyone currently working on it
42
Intercognitive retweeted
Learn more and book your pilot at auki.com/getcactus
We've built most advanced AI system for retailers in the world. Now with autonomous robots, AR navigation and an AI copilot that helps you run the store.
11
39
1,379
Who would be robotics canon?
3
1,698
Go outside! Have the robots touch grass!
Replying to @NVIDIARobotics
Simulation gives the brain a controlled place to learn, sub-centimeter RTK gives the body the ground positioning it needs once it steps outside the lab. Bridging sim-to-real for outdoor Physical AI. 🤖
1
6
241
Let a thousand robot-folded origami cranes fly
Sharpa will be co-hosting two competitions at IROS 2026. Powered by Sharpa's robot platforms and software, the 2nd ROCO and Origami challenges (cohosted by @BitRobotNetwork) brought together 48 teams from leading institutions around the world—including Carnegie Mellon University, UC Berkeley, KAIST and NUS. Across both competitions, teams will put robotic manipulation and dexterity to the test on Sharpa hardware through challenging real-world tasks. Follow the challenges for more: ROCO: @NtuLab38456 Origami: robotic-origami-challenge.gi…
60
Intercognitive retweeted
For those who missed it yesterday, I wrote the ACT explainer I wish I'd had when I got into robot learning. For ML people, no robotics background assumed, with animations. What's in it: - what goes wrong when predicting one action at a time - how predicting action chunks helps - why regressing on good demos can give you a bad policy (clip below) - what ACT's latent is for, and whether it matters Next one will be on Diffusion Policy. Follow me to catch it.
Instead of training to predict the next action, ACT (@tonyzzhao et al.) trains to generate the next 100. How and why, explained for ML people new to robot learning. Second from my speedrun. Next one, on Diffusion Policy, next week. Follow me to catch it.
Article

How ACT works

If you're an ML practitioner entering robot learning, you'll likely train your first policy with ACT. My hope is that this article will make every piece of it feel natural. I'll cover what ACT is, why

1
16
141
16,038