Video: Robot walks into 30 stranger homes — then makes beds, folds towels and cleans up

A humanoid robot enters a home it has never seen. No map. No special training. No prior data from inside the house.

Then it starts making the bed, folding towels and tidying the living room.

California-based robotics company Figure has unveiled Helix 2.5, a new artificial neural network (ANN) that it says allowed its humanoid robot to complete household work across 30 unfamiliar homes without collecting data in those properties or fine-tuning the system for their layouts, furniture or objects.

Figure said the robot was tested in 30 previously unseen homes in the San Francisco Bay Area.

The homes were excluded from the model’s training process, and the company said it used a fixed version of the system for each task rather than adapting it from house to house.

The robot was assigned three long-horizon, whole-body household tasks:

Figure said Helix 2.5 could navigate different layouts, handle unfamiliar furniture and objects, and recover after mistakes without direct human intervention.

The company described the work as “zero-shot” generalisation — meaning the robot had not been trained or adjusted using data from the homes or the specific objects used in the tests.

Robots have performed impressive demonstrations for years. But most operate in tightly controlled settings: factories with fixed workstations, warehouses with standardized shelves or carefully prepared lab environments.

A real home is harder.

Furniture is placed differently. Towels vary in size and material. Beds sit at different heights.

Objects are cluttered, partially hidden or positioned in ways the robot did not expect. Lighting, flooring and room dimensions also change.

Figure’s claim matters: for a home robot to be useful, it must do more than repeat a memorised task. It must recognise a new environment, understand a verbal or visual goal, use its arms, hands and body safely, and adjust when the first attempt does not work.

Figure said Helix 2.5 was pretrained using its proprietary "Index dataset" of human behaviour.

In a controlled comparison, the company said pretraining raised zero-shot whole-task success to 56% from 9% for an otherwise comparable model trained from scratch, according to The AI Insider.

The reported results are company figures and have not yet been independently peer-reviewed.

The demonstrations point to rapid progress in embodied AI: artificial intelligence that does not only generate words or images, but perceives and acts in the physical world.

Still, there is a large gap between a successful test and a commercially reliable household robot.

A product used daily in homes would need to work safely around children, pets, stairs, fragile items, hot appliances and unpredictable human behaviour. It would also need long battery life, low maintenance, robust privacy protections and a price that ordinary households can afford.

The robot would have to complete tasks close to perfectly, not merely succeed in a majority of trial runs.

Figure is part of an intensifying race to build humanoid robots that can work in warehouses, factories, logistics centers and, eventually, homes.

Companies are betting that human-like bodies offer a practical advantage because the built world — doors, stairs, tools, kitchens and workplaces — was designed for people.

For now, Figure’s Helix 2.5 is a research milestone, not a consumer robot on sale at the mall.

But the image is hard to ignore: a machine walking into a stranger’s home and beginning to work without being shown where anything is.

The long-promised household humanoid may not have arrived yet — but it is starting to look less like science fiction.