Twitxy Saved

Continuum Labs

Verified account This account is verified. Learn more

@continuumlabs_

Trust layer for Physical AI.

Discord: discord.gg/uvhrtzCqv
Telegram: t.me/+Nm2xhKaU9wFiY…
World continuumlabs.one Joined July 2026
8 Following 2.1K Followers 2K Posts

Latest posts 5

Continuum Labs Verified account @continuumlabs_
How long can a robot actually work before it needs to recharge?

It depends on what the robot is doing.

Walking, lifting objects, processing camera feeds and running AI models all consume energy. A robot performing demanding physical tasks may drain its battery much faster than one handling lighter workloads.

And the battery isn't the only factor. Powerful onboard computers, sensors, motors and cooling systems all compete for the same energy supply.

This creates an important challenge for autonomous machines: balancing performance, compute and energy consumption without interrupting operations.
Continuum Labs Verified account @continuumlabs_
Physical AI this week was about models that actually move robots and capital that follows real work: open world-action models, Gemini Robotics 2.5 on the factory floor, and two big manufacturing rounds.

Here's what made the headlines this week:

1️⃣ @nvidia introduced GR00T 2, an open System One model that learns skills from human demos and transfers them across different robot bodies.

2️⃣ @GoogleDeepMind released Gemini Robotics 2.5, physical AI that can learn hundreds of skills, from folding laundry to sorting car parts on a production line.

3️⃣ @runwayml open-sourced Praxis-1, a 27B world-action model under Apache 2.0 that turns video pretraining into robot control.

4️⃣ @PathRobotics raised a $300M Series E at a $2.45B valuation, led by @Georgian_io, to expand physical AI beyond welding.

5️⃣ Agile Robots closed a $500M Series F at an $8.5B valuation, led by SoftBank, to expand U.S. operations and production capacity.

6️⃣ @GXOLogistics signed a letter of intent with @Apptronik to expand Apollo trials and potentially deploy up to 1,000 humanoids over four years under a RaaS model.

7️⃣ Swiss startup Embodied AI unveiled AIR, a wheeled humanoid with tendon-driven arms designed to learn factory tasks from workers and work beside people without cages.

Follow @continuumlabs_ for more weekly updates like this.
Continuum Labs Verified account @continuumlabs_
Why do so many robots have two arms, two legs, and a human-like body?

Because much of the world is already designed around us.

From door handles and staircases to tools, workstations, and vehicles, our infrastructure assumes a human body will interact with it.

Humanoid robots could use many of these existing spaces and objects without requiring major changes to the environment.

But copying the human form doesn't automatically give a robot human capabilities. It still needs to learn how to move, perceive, manipulate objects and respond to unfamiliar situations.

Humanoids may become a familiar part of the machine economy, alongside specialized robots built for specific tasks.
Continuum Labs Verified account @continuumlabs_
Opening a door seems simple until you ask a robot to do it.

A robot needs to locate the handle, understand its shape, position its hand correctly, apply enough force to turn it, and coordinate that movement with the door itself.

A small change in handle design, resistance, or door weight can make the task much harder. The robot must constantly use visual and force feedback to adjust its movements.

It's a good example of why everyday tasks remain challenging for robotics. Physical intelligence requires more than knowing what to do.

A robot needs to interact with the world and adapt to what happens next.
Continuum Labs Verified account @continuumlabs_
A robot can perform a backflip, but loading a dishwasher requires a different kind of intelligence.

The difference is predictability.

A backflip can be learned as a sequence of movements and repeated under familiar conditions.

Loading a dishwasher means dealing with objects of different shapes, sizes and materials, figuring out how to grip them, where they belong and how to avoid breaking anything.

Every small change in position creates a new problem for the robot to solve.

This is why everyday manipulation remains one of the hardest challenges in robotics.

Robots need to perceive their surroundings, plan movements, and constantly adapt to what's happening around them.