UCLA builds a humanoid from smartphone

PLUS: Tesla is developing a robot skin

AuthorAuthorAuthorAuthor

Jennifer Mossalgue, Rowan Cheung, Shubham Sharma & Zach Mink
October 12, 2026

Good morning, robotics enthusiasts. That Android smartphone gathering dust in your junk drawer could have a second life.

UCLA researchers built PhoneBot, a 19-inch humanoid with a smartphone for a brain and $400 in extra hardware. It walks, twists its torso, and pushes itself back up after a faceplant.

Also: Tesla’s patent application details touch-sensitive robot skin, while EPFL unveils a robot motor that looks like thread.

In today’s robotics rundown:

  • UCLA turns old phones into mini-humanoids

  • Tesla seeks patent on robot skin

  • Robots hit the gym to learn real work

  • Tiny motors could give clothes muscle

LATEST DEVELOPMENTS

UCLA

Image source: Images 2.5 / The Rundown

The Rundown: Your retired Android smartphone now has a new use case: UCLA’s PhoneBot puts a phone in charge of a miniature humanoid with an estimated $400 hardware bill — phone sold separately.

The details:

  • PhoneBot stands ~19 in. tall and uses 13 motors to walk, swivel its torso, and push itself upright from a face-down position.

  • Four Android phones handled its core sensing, vision, and movement control, including a 2017 Honor 9 that lacked the services needed for onboard mapping.

  • Researchers corrected an uneven gait by adding mirrored training examples that swapped left and right, helping the controller learn more balanced walking.

  • Still, the PhoneBot has no arms, has only been trained to walk on flat ground, and needs a separate laptop for speech recognition and its language model.

Why it matters: OpenBot already paired Android phones with $50 wheeled bodies in 2020; PhoneBot extends the idea to the trickier business of balancing on legs. While it has room to grow, this scrappy little machine could give students and small labs a foothold in humanoid robotics.

TESLA

Image source: Tesla / USPTO

The Rundown: Tesla just applied for a patent on a flexible touch-sensing skin that can be shaped to fit curved surfaces such as a robot’s thumb, palm, and fingers, according to a filing with the USPTO — and its hand is where Optimus has struggled most.

The details:

  • Tesla filed the patent application on Sept. 11, 2025, and it was published on Oct. 8, 2026; it has not been granted.

  • The sensors sandwich a pressure-sensitive material between flexible layers, translating contact into electrical signals across curved fingertips and palms.

  • Tesla describes printing sensor patterns onto flat materials before molding the assembly into shape, offering a potential route to tactile “skin” at scale.

  • The application’s “3D” terminology refers to the sensors’ curved geometry; the filing doesn’t establish three-axis force sensing or confirm deployment in Optimus.

Why it matters: The Information reported last month that hand assembly, supplier issues, and equipment failures — including fragile touch sensors — threaten Tesla’s 1K-robots-a-week target by year-end. The filing doesn’t say whether this design fixes any of that, and Optimus V3’s public debut is already delayed.

ROBOT DATA

Image source: Mecka AI

The Rundown: Robotics companies are spending billions to teach machines something humans take for granted — how to do everyday tasks — by recording people at work and at home, the Financial Times reports. 

The details:

  • Robotics firms are building “robot gyms” and filming everyday tasks to fill a shortage of real-world training data.

  • Figure pays ~44K participants an average of 94 cents per task and plans to spend over $1B generating data over 12 months.

  • Robotics and physical-AI companies have raised nearly $48B this year, according to PitchBook figures cited by the FT.

  • Data supplier Mecka raised $60M in a Sequoia-led round backed by Nvidia to expand its data collection and robot deployment business.

Why it matters: Getting robots to work reliably in unfamiliar homes and workplaces means teaching them to handle situations they haven’t seen before. Companies are betting billions that enough human demonstrations will build that adaptability, but it’s still unclear whether more data will deliver more capable machines.

EPFL

Image source: EPFL

The Rundown: Your next wearable robot could be hiding in the seams: EPFL’s Soft Transducers Laboratory researchers built a flexible, threadlike motor that could put movement directly into clothing.

The details:

  • The FiberMotor nests one fiber inside another, with coiled electrodes using electrostatic attraction to drive movement in either direction.

  • Four bundled motors lifted a 46g chocolate bar and flexed a robotic finger; individual motors held stationary loads equivalent to 75g.

  • It’s backdrivable — if a wearer moves against it, the fibers slide past each other. The team has already stitched it into a prototype garment shaped to fit a knee.

Why it matters: FiberMotor could distribute force across clothing without the bulky gears and transmissions often used in wearable robotics. The team is working to make it stronger and more durable as it pursues commercialization for soft exosuits, VR haptics, and lightweight prosthetics.

QUICK HITS

Germany’s Agile Robots is using videos of people performing tasks, combined with simulated practice, to overcome the shortage of training data, Bloomberg reports.

IHMC’s Alex humanoid is in U.S. military field tests for hazardous tasks, but it reportedly still needs a remote operator and stumbles on basic actions.

Uber and China’s Pony.ai plan to begin testing robotaxis in London within weeks, expanding their European partnership, with no date yet for public rides.

XPENG plans to launch its IRON humanoid in Europe in the second half of 2027, initially for commercial tasks such as guided tours and shopping assistance.

BYD’s new patent images show its proposed humanoid design, but the company hasn’t confirmed specs or production plans.

Brooklyn-based Ultra raised $62M and partnered with Physical Intelligence to scale its AI warehouse robots, which have already fulfilled 500K+ orders.

AthenaZero, from the Robotics and AI Institute, uses low-inertia arms and yielding joints to throw, catch, and hit balls at speed.

Food delivery robots across U.S. cities are being kicked, overturned, and smashed, reflecting both casual vandalism and broader resentment toward AI, the NYT reports.

Meteomatics said its reusable Strato drone reached 34,245 ft. (10,438 m) to collect weather data, offering an on-demand alternative to weather balloons.

San Diego-based Psionic showcased its 490g bionic robotic hand with 30 touch sensors and six motors, capable of delicate tasks like holding a raspberry.

That's it for today!

Before you go we'd love to know what you thought of today's newsletter to help us improve The Rundown experience for you.

Login or Subscribe to participate in polls.

See you soon,

Rowan, Zach, Shubham, Jennifer, and Nate — The Rundown’s editorial team