The Neo Home Robot: The Tesla Playbook for Embodied AI
Neo’s often tele-operated ‘AI’ robot isn’t completely autonomous yet, but this data-first strategy could make it unstoppable.
With much buzz and excitement ‘1X’ has put their much anticipated humanoid home assistant bot ‘Neo’ up for pre-order. For the princely sum of $500 monthly or a $20,000 one time fee, Neo promises to be your perfect house keeper, completing all your chores whilst you work, allowing you to return to a clean and orderly home, all thanks to Neo.
But despite the marketing material, it’s seems that at the present moment, the undoubtedly impressive bot is more of a list of promises and aspirations than complete product.
In his video titled “The Problem with this Humanoid Robot”, Marques Brownlee highlights all the ways the dream and the reality aren’t meeting up quite yet. With much of the demo’s given in their marketing material and to journalists being tele-operated by human “pilots” using VR goggles and controllers. If you haven’t seen the video i recommend you check it out here:
Whilst at first light this might seem like a “scam” or a simple cases of false advertising, the reality is far more interesting. Marques suggests that 1X is reading from the Tesla self driving playbook and that building a fleet of tele-operated bots (what they call expert mode) now will yield huge dividends in the long term.
Looking at the hardware it’s self we see that this first generation Neo bot is no toy. It’s a serious piece of hardware and engineering, with my guess being that the $20,000 price tag is far below current manufacturing costs. It has dexterous hands capable of manipulating objects, it’s capable of lifting over 150 pounds (68 kg) and carrying 55 pounds (25kg), has a continuous working battery life of 4 hours and is boast the latest NVIDIA Jetson Thor chip designed specific for embodied AI applications.
Redwood AI is 1X’s in house AI model that drives the brains of Neo. But despite all the impressive progress made with their in house VLA model the software is not quite ready for prime time as explained in the Marques Brownlee video.
The issue is foundational models such as Redwood AI, is that they need and crave only one thing:
Data.
Tesla understood this fundamental problem when it set out to build a generalised solution to self driving. Cameras on every car they sold helped to train successive generations of better and better FSD models. FSD 14 appearing to be moment that Tesla can begin offering self driving taxi services, a key Tesla growth area.
The clever thing about this approach is that Neo doesn’t need to be perfect yet. Every moment it’s being operated, every object it picks up, and every mistake it makes is valuable training data. 1X isn’t just selling a robot, they’re deploying a fleet of data-gathering machines into real homes. As Marques Brownlee points out, those tele-operators using VR headsets aren’t just filling in for missing intelligence, they’re actively teaching Redwood AI how to perceive, react, and move in the real world.
Tesla understand that data is one of their biggest advantages. 1X seems to be following the same playbook, only this time the training ground isn’t public roads but the private, messy interiors of our homes. If they can get enough bots out there, each one becomes a data node in a network that helps the entire system learn faster. The long-term value of that data could be enormous.
Of course, this all comes with a cost. Those early adopters are paying to be part of the experiment. For $20,000 you’re not just buying a household helper, you’re helping train one. Privacy will be the first concern, as these robots will map homes, record actions, and inevitably see things most people don’t want seen. 1X says they’ll blur faces and allow geofencing, but inviting cameras, microphones, and remote operators into your living room is still a big ask. The trade-off for being early is not just money, but trust.
In the end, Neo isn’t really about washing dishes or folding laundry, not yet anyway. It’s about building the foundation for embodied AI, a system that can think and act in the physical world as fluidly as language models handle text. 1X is selling the dream early to build the future faster, and while that might frustrate some, it’s also a bold and probably necessary step.
Whether Neo becomes the next big platform or just another footnote in the AI timeline will depend on how fast they can turn that data into real capability. Either way, it feels like the start of something important.