When 1x Technologies unveiled the Neo robot last week, the demonstration footage looked impressive: a lithe humanoid folding laundry, clearing dishes, moving through a home with apparent autonomy. Then the Wall Street Journal revealed the uncomfortable truth. Nearly all of those tasks were puppeteered by a human wearing VR goggles, operating the robot's limbs remotely. The company, founded in 2014 by roboticist Björn Þórisson and now backed by OpenAI's Equative Ventures, isn't hiding this anymore—they're leaning into it.
This might seem like a setback, but the architecture suggests something more calculated. Neo is being offered through two payment models: $500 monthly subscription or a $20,000 one-time fee, with a $200 deposit to reserve. The robot itself is deliberately non-threatening—soft-clothed, weak-limbed, with cartoonish eyes. It can't hurt you if it malfunctions. This design choice, combined with the transparent acknowledgment of human operation, serves a dual purpose. First, it makes the technology acceptable: people fear the Optimus bot's industrial precision; they might tolerate Neo's bumbling clumsiness. Second, and more importantly, it establishes a pipeline for something far more valuable than a domestic helper.
The real play here is data. Large language models were trained on centuries of written text freely available online. But visual and spatial data—how humans actually move through rooms, manipulate objects, interact with appliances—is scarce. Cameras are recent. Home sensor data barely exists. A Tesla collects dashcam footage to train self-driving models; Amazon's smart home devices capture architectural layouts but nothing rich enough to train humanoids. But a robot controlled by a human, recording from multiple angles while operating in the chaotic, variable environment of a real home? That's the training gold mine. Each household is different. Every kitchen layout varies. Every person stacks dishes differently. For an AI system learning to handle unstructured environments, this is incomparably richer than factory floors or controlled industrial settings.
The company is essentially running a distributed data collection operation disguised as a service business. Early customers pay to have their homes recorded by remote operators, generating the raw material to train the next generation of autonomous systems. Once Neo has enough recordings—thousands of homes, millions of manipulation sequences—the autonomous version arrives and the human operators become unnecessary. The model becomes efficient. The subscription converts to pure revenue. This is the Tesla playbook applied to domestic robotics: give away the data collection layer, train the model at scale, launch the autonomous product later.
What makes this clever is the sequencing. They're not pretending autonomy they don't have. They're not overpromising timelines. They're pricing the service as a hybrid product—human + robot—which gives it immediate viability while establishing customer acceptance. And they're doing it quietly, betting that most people won't notice or care who's actually controlling their home robot, as long as the dishes get done.