Signals Inbox·July 16, 2026·Humanoid Robotics

Humanoids: what household tasks can they do today?

Humanoids can already load dishwashers, collect toys, fold towels, wipe surfaces and reset prepared rooms. The impressive part is how quickly these sequences are getting longer. The less impressive part is that ordinary homes are still far too unpredictable for robots to manage without regular human help.

Humanoid household tasks ranked by current capability

Q1Which household chores are closest to becoming genuinely useful now?

Dishwasher handling, basic room tidying and towel folding are the household chores closest to becoming useful.

Only nine of the 30 tasks in our tracker score above 55, and seven come from Figure’s kitchen, living-room or bedroom demonstrations. Domestic humanoid progress therefore depends heavily on one company and a few prepared environments. Independent replication remains limited.

The leading tasks all reduce uncertainty. Dishes begin inside a known appliance. Toys go into a designated container. Towels are placed on a table. The robot performs the physical sequence while the homeowner still defines what needs doing and where everything belongs.

Figure has nevertheless lengthened its autonomous sequences quickly. It progressed from single-station towel and dishwasher tasks in 2025 to full-room kitchen, living-room and bedroom demonstrations in 2026. That is real progress, but measured in minutes rather than hours of household work.

Dishwasher handling ranks first because it combines fragile objects, walking, storage and a complete sequence. Collecting visible objects into predefined containers follows. Towel folding is credible, but far less general than processing an ordinary basket of mixed clothes.

Q2Can a humanoid clean an entire home autonomously today?

No humanoid can clean an ordinary home autonomously today.

Figure’s strongest public sequences cover a dishwasher, one living room and one prepared bedroom. Together, they amount to less than ten minutes of published whole-room activity. Bathrooms, floor cleaning, food waste, mixed laundry, spills and restocking remain outside that chain.

Longer tasks also multiply failures. A robot that succeeds on each individual action 95% of the time has only about a 60% chance of completing ten such actions consecutively without an error. At 90% reliability per action, that falls to roughly 35%. A useful whole-home routine may contain hundreds of actions.

Research outside humanoid companies shows the same problem. Dobb-E achieved an 81% success rate across learned tasks in real homes after receiving demonstrations for each one. That is strong research performance, but it still implies approximately one failure in every five attempts, before adding the balance challenges of a biped.

The realistic near-term product is a robot that completes selected routines in a familiar home: unload the dishwasher, collect visible toys and fold a prepared towel pile. “Clean the house” remains far too open-ended.

Q3Are home humanoid robots autonomous or still remotely controlled?

Home humanoids currently mix autonomous execution with remote human assistance.

Figure’s recent household videos are presented as autonomous, including its four-minute kitchen sequence without resets or intervention. 1X is taking a more commercial approach: NEO will perform basic tasks independently, while remote experts can supervise or control unfamiliar chores.

Its cooking demonstration shows the difference clearly. NEO Beta appeared to prepare a steak, but every movement was controlled by a person using a virtual-reality headset. Another person activated the burner. The robot proved that its body could perform the motions. It did not prove that the robot could cook.

Remote control also creates training data. Human operators demonstrate tasks, the robot learns from those examples, and later versions may execute parts of the same workflow independently. Teleoperation works both as a fallback service and as a development method.

The most useful missing metric is the remote-intervention rate. One intervention every few weeks would feel autonomous to an owner. One intervention per chore would make the product closer to a physical telepresence service. No major developer currently publishes this figure.

Q4What will be the first household task people genuinely delegate to humanoids?

The first household task people regularly delegate to humanoids will probably be resetting rooms by collecting visible objects and returning them to predefined places.

We assign roughly 45% probability to room resetting, 30% to dishwasher handling, 20% to simple laundry folding and 5% to other tasks.

Tidying is broad enough to create an obvious result but can still follow simple household rules: toys go into one bin, clothes into another, pillows return to the couch and chairs go under desks. Figure has already demonstrated most of those actions, while 1X is designing NEO around scheduled household routines.

Dishwasher handling currently has stronger autonomous evidence, but kitchens add broken glass, sharp utensils, food residue and different storage conventions. Room tidying usually involves lighter and safer objects.

Humanoids will also struggle to compete with specialized appliances at the tasks those machines already perform well. Robot vacuums clean floors, dishwashers wash dishes and washing machines wash clothes. The useful gap lies between them: collecting objects, loading machines and returning items afterward.

Confidence remains moderate because no developer has disclosed sustained consumer usage. Early delegation may also rely on remote assistance. Still, room resetting has the clearest combination of technical feasibility, low physical risk and visible household value.