Architecture Versus Glove Data Advantage

This comes down to whether better priors or more household reps become the tighter bottleneck first. 1X is betting that a humanoid body plus a world model lets NEO reason through unfamiliar cabinets, stairs, and cluttered rooms with less task specific training, while Sunday is betting that thousands of glove demonstrations across 500 plus homes let it learn the small hand motions and object handling details that actually make chores work in real kitchens and laundry rooms.

• 1X has built a clear architecture loop. NEO is shaped for human spaces, can be remotely guided when autonomy fails, and 1X says its world model improves as training compute and real world NEO data increase. That favors faster transfer from one task to the next once the base system is good enough.
  2 sacra 3 1x 4 1x
• Sunday has built a clear data loop. Its glove is designed to mirror Memo’s hand and sensor layout, and the company says it has already distributed 1,000 plus gloves across 500 plus households before broad robot deployment. That can create a larger corpus of real home manipulation data earlier and more cheaply than shipping full robots.
  1 sacra 5 sunday 6 sunday
• The practical split is breadth versus density. 1X can target the whole home because a biped can open doors, climb stairs, and use cabinets as they already exist. Sunday can focus on higher frequency chores at lower cost, with a wheeled base and sub $10K target, but it needs its data advantage to offset the limits of a less human shaped machine.
  1 sacra 2 sacra 5 sunday 6 sunday

The likely next phase is convergence. World models become more valuable as real home data piles up, and large household datasets become more valuable when paired with robots that can generalize beyond scripted motions. The company that closes that loop fastest will set the pace for home robotics, because the winner is the one that learns new chores in days instead of product cycles.