September 25, 2026
Scan today's general-robotics headlines and you won't find a single new humanoid or bipedal platform. IEEE Spectrum covered a student team in Guadalajara, Mexico building a portable educational robotics kit for high schools that lack lab access. Robohub ran an interview with Hello Robot co-founder Aaron Edsinger about the company's Stretch robot, which helps people with mobility issues live more independently. Robohub also covered Intrinsic -- Google's AI robotics group -- open-sourcing parts of its ROS-compatible platform, called Intrinsic Core, and previewed the IROS 2026 conference program set for Pittsburgh next September. None of these are about humanoids in the torso-arms-legs, walk-and-manipulate sense our beat usually tracks.
That's worth stating plainly rather than papering over. We could stretch Hello Robot's Stretch platform to fit -- it does have an arm and a mobile base, and Edsinger himself is a veteran of humanoid research going back to MIT's Domo project -- but Stretch is explicitly a wheeled, single-arm assistive robot, not a bipedal humanoid. Calling it one would be the kind of category-blurring that makes robotics coverage less useful, not more.
The more interesting story today is actually about infrastructure, not embodiment. Intrinsic opening up Intrinsic Core suggests Google believes the near-term battleground in "physical AI" isn't which company ships the flashiest humanoid demo -- it's who controls the software layer that any robot, humanoid or otherwise, runs on. If ROS-compatible, open capabilities from a Google-backed player gain traction, it lowers the barrier for smaller robotics startups to build sophisticated applications without reinventing control stacks from scratch. That's a slower, less visible kind of disruption than a viral walking-robot video, but it's arguably more consequential for who actually wins market share over the next five years.
Hello Robot's pitch is a useful reality check too. Edsinger's team has spent years building toward practical, safe, affordable robots for eldercare and mobility assistance -- a market that's enormous and underserved, but one that rewards reliability over spectacle. It's a reminder that the humanoid hype cycle, dominated by headlines about warehouse and manufacturing deployments, can obscure quieter, arguably more human-centered applications that don't need two legs to matter. We've made a similar point before about China's humanoid roadmap being fundamentally a manufacturing and industrial-policy play rather than a sci-fi spectacle -- the same logic applies here in reverse: not every useful robot needs to look like a person.
If you're a business leader tracking robotics for eventual deployment, days like today are actually the more honest signal than the demo-heavy weeks. Development pipelines -- open-source tooling, education programs building the next generation of engineers, conferences like IROS 2026 where research actually gets vetted -- are the unglamorous scaffolding that determines whether the humanoid promises made in 2024 and 2025 turn into deployable products in 2027 and beyond. A steady diet of viral demos without this kind of underlying infrastructure work is usually a warning sign, not a good one. Just as we've argued that ungoverned code creates a bill that comes due later in software, under-invested robotics infrastructure creates the same deferred cost in hardware -- it just takes longer to show up.
What would change your mind about how close humanoid robots really are to your industry -- a manufacturing deployment number, a price point, or something else entirely?
Sources