Martinez, California, is as far as you can get from Silicon Valley and still in the San Francisco Bay Area. Perched on the northeastern edge of the bay, the small town is home to Hello Robot, a start-up in itself that’s about as far from the big promises of its robot rivals 45 kilometers to the south.
Hello Robot released the fourth iteration of its home assistance robot, Stretch, last month. And you might stretch to call it a humanoid robot. While Stretch boasts a humanoid body and sensor-filled head, its telescoping arm has pins, and it rides around on a heavy, omnidirectional wheeled base.
When Stretch’s batteries run out, the lights around its “eyes” light up — “it looks angry,” Blaine Matulevich, an engineer at the company, jokes.
Hello Robot, founded in 2017 by CEO Aaron Edsinger, former director of robotics at Google, and CTO Charlie Kemp, a professor at the Georgia Institute of Technology, doesn’t build a basic model or promise to take over every job a human can do. Hello Robot has been developed to expand to do something most other robots don’t: Work in real homes, with real people, at a time when most are behind glass in laboratories.
This is important. Although recent advances in artificial intelligence promise more power for robots, there is a lack of useful training data. And while simulations are improving, investors are increasingly focused on implementation.
“First-mover companies are accumulating domain-specific recovery loops and workflow tolerances that no competitor can buy or integrate,” Bullhound Capital wrote in an industry report published last week. “For robotics, the moat is not just the IP, but the hours of work accumulated under real-world responsibility.”
A different kind of illustration
Keith Platt, an investor from Georgia who now sits on Hello Robot’s board, invested in the company after taking Stretch as a roommate. Platt became a quadriplegic in 2021, only able to control parts of his shoulders, his neck and his head. He began exploring adaptive technology, and in 2024 began working with Hello Robot, which has an occupational therapist on the team to support its work with Platt and other people with similar conditions.
Platt controls his Stretch using a voice-activated iPhone app; he can challenge you to move to a certain place in his house, and take direct control to manage things and perform tasks. Another deceptively simple project has been finding a way to get Stretch to give him a protein shake for breakfast, which usually requires someone else’s help.
“When we started that project, it took me independently — no one there — it took about two hours,” Platt told TechCrunch. “But I’d keep holding on to it. It got there, within a few minutes, and I’d drink the whole shake and put it back on the counter.”
Being dependent on people is a real challenge, both physically and emotionally, Platt said. Anything she can do to regain her independence — like putting on or taking off her reading glasses, or brushing herself — is “big.” Not only for him, but also for the people who care about him.
He predicts that it will be “life-changing” for families if robotic assistants can enable people with mobility challenges to spend the day safely at home, allow their family members to work independently or leave the house without hiring a professional caregiver.
Stretching comes from a factory with limited autonomy; focusing on having someone in the loop is the goal. “Control is a factor — it’s desirable to include a robot,” Matulevitch said.
And, Platt points out, he’s not worried about Stretch falling over if it goes wrong.
The hardware is solid
For all the money flowing into startups designing robot brains, their bodies still leave a lot to be desired. Although the parts are cheap, the state of the art still delivers heavy parts that require high strength, efficient balancing. A robotic hand and arm weighs more than a human, and physics is not forgiving.
When robots make mistakes, they damage the things around them. The first, the Bot Company, is being sued by an Airbnb owner in San Francisco who claims the company rented his space to work on its robot, which scratches furniture, breaks appliances, and tiles bathrooms.
“The state of hardware today is actually pretty bad from the point of view of, ‘I want to have robots in my parents’ place,'” Mahi Shafiullah, a robotics postdoc at the University of California, Berkeley, told TechCrunch. He recalled industrial robots in his lab accidentally puncturing a plastic kitchen toy he was supposed to carefully manipulate.
Shafiullah ended up using the third generation of Hello Robot’s Stretch as part of his PhD research at New York University. The models he helped develop with Stretch won the best of presentation award at least at the annual Computer Vision And Pattern Recognition (CVPR) conference.
Hello Robot doesn’t promise the Stretch will have the sophistication or power of humanoid robots that appeal to Valley, but its simple design could make it very powerful. Edsinger compares his company to Waymo, which became a leading innovator of self-driving cars by focusing on safety first (though money helped).
One leader in this field, 1X, became the subject of significant attention last year when it introduced a humanoid robot, Neo, that people can buy to do household chores. The company says it has sold 10,000 Neos it plans to build this year, but so far, none have been delivered.
“Hello, the robot was really careful and really cared about this problem, because I think they designed it to be among people first,” said Shafiullah. “And then they think, what skills can they fit into those limitations?”
I’m heading home
The Stretch 4 costs $30,000 for an affordable robot, which is slightly more than robots from Chinese manufacturers, though Edsinger notes that those often don’t come with sensors or software installed, additions that ultimately drive up the price. He expects to produce between 200 and 300 at the company’s Martinez headquarters, and the first one is already finished.
Edsinger wants to keep the robot accessible to hackers and researchers with low budgets. One design flaw of Stretch is that it must be shipped in a cardboard box via UPS or DHL—if wooden crates and installation teams are required, costs increase and accessibility decreases.
Hello Robot Customers include researchers using Stretch to test developing AI brains, business customers exploring the use of Stretch in settings such as data centers, and people working to develop home assistants for people with disabilities.
The combination of the robot’s sensor suite, physical strength, and safe operation could make it a candidate to fulfill the hopes of physical AI believers.
“Algorithms may be there, but data is not, and data is actually like 80% of the important ingredient,” Shafiullah said.
Having a robot that can safely collect that data is another step forward. And Hello Robot intends to keep repeating. Lessons learned from the release of Stretch 4 promise to feed into the company’s next bot, which could lower the price and increase the power enough to realize the vision of robot interaction with people in the home.
If you shop through links in our articles, we may earn a small commission. This does not affect our editorial independence.
