Apple wants to build a home robot, which is adorable because the company still hasn’t figured out how to make Siri understand a basic request without melting down. A robot is basically Siri with legs, and right now, Siri can barely handle a grocery list.
That’s the context for Apple’s latest leadership change in AI. John Giannandrea is moving into an advisory role after years of trying to clean up the company’s scattered AI strategy.
His replacement, Balamurugan Subramanya, is inheriting a very Apple problem. Apple keeps delivering great hardware and monster chips, and the presentation is always immaculate. The problem is that the intelligence holding everything together still feels like it never made it past 2018.
And that’s before we even get to the robot idea. Real robots need real awareness. They have to map rooms, plan movements, and respond to whatever you throw at them, including the slightly different way you ask for the same thing every day.
Apple has been trying to fix Siri for over a decade. Meanwhile, Amazon can barely get Astro to patrol a hallway without bumping into a chair. Robotics is unforgiving. If your AI is confused, your robot walks into a wall.
Apple is also trying to build this in a privacy-forward structure that famously avoids collecting the kind of data these systems thrive on. It’s the right choice ethically. It’s an incredibly hard choice technically.
A robot that lives in your home needs to understand your home. That’s not something you solve with a couple of on-device models and a keynote montage of people smiling in modern kitchens.
The bigger question is whether Apple can actually commit to the messy iteration robotics requires. This is not a category that rewards perfectionism. You learn by shipping, failing, and shipping again. Apple’s culture rewards polish, not public flailing.
If Apple really wants a home robot, it needs to prove it can deliver day-to-day intelligence on the devices people already own. Siri should feel reliable before anything starts rolling around your floor.
Otherwise, the whole thing risks turning into a slick machine that shows off Apple’s engineering chops while quietly proving that the real problem was the intelligence all along.
