Robots at the big tech expo never fail to offer both wonder and whiplash. This year’s show floor brought a selection of three that were part practical, part bizarre, and indicated both where home robot devices are going and why they still seem so otherworldly.
These aren’t factory workhorses. They’re I/O-binding types, meant for your living room or laundry room, where performance sensitivity is equal to safety, personality, and price. Here’s a closer look at the trio causing the most double-takes — and conversation.
LG’s CLOiD Robot Has Humanlike Hands to Help With Laundry
LG’s CLOiD concept embodies the dream of a robot housemate: two arms with five-digit hands, a voice that sounds just like any other person’s, and a vaguely humanoid body.
On the show floor, it talked, maneuvered, and, most provocatively for those who believe robots will someday make everything, folded laundry — a deceptively difficult test because cloth is floppy, shapeless, and unforgiving of clumsy grippers.
History is simply littered with laundry robots that never emerged from the lab or the purgatory of crowdfunding. Remember FoldiMate and Laundroid? Both demonstrated just how rough it can be to work with bendable objects in the wild. Grasping has seen improvements from research at universities and labs, with innovations like vision-language policies and diffusion-based control, but transforming demos into reliable home routines is leaping across a gap. If CLOiD progresses, questions about safety certifications like UL 3300 for service robots, pinch-force limits around hands, and whether perception runs on-device versus in the cloud will follow.
And yet, the ambition landing zone of CLOiD is rather sweet: It’s the boring stuff that people actually do want some help with. If it can fold, sort, and shuttle laundry consistently, though, it sidesteps the “gadget” trap and edges toward a useful assistant.
Takway Sweekar Turns Generative AI Into A Shape‑Shifting Companion
The strangest of the lot is Sweekar, which questions what counts as a “robot,” anyway. You start with an egg, and through nurturing, it becomes a personality shaped by generative AI. Think Tamagotchi for the LLM era: care loops, growth stages, and customization — probably a little less task robot, a lot more digital pet with actual heft.
The pitch addresses a genuine need. The U.S. Surgeon General has described loneliness as an epidemic, and robot companions — ranging from Sony’s Aibo to therapeutic social bots for elder care that have actually been measured for their positive impact — are known to benefit at least some users. But AI as a companion throws up difficult issues. What data trains the personality? How are conversations stored? What protections resemble requirements such as COPPA if kids are in the equation? Ethicists and organizations such as the IEEE Standards Association have cautioned against over-anthropomorphizing machines; its design, like Aldebaran’s, will require transparent guardrails as much as charm.
If the engagement loop does work — daily rituals, visible growth, and a sense of co-creation — Sweekar could launch a brand-new consumer category that combines game mechanics with embodied AI.
SwitchBot Onero H1 Gambles On Mobile Manipulation At Home
Onero H1 resembles a tall, rolling post with two delicate arms and a head — more practice dummy than humanoid. The form factor, however, is a savvy midpoint: a mobile bottom to tote from room to room, arms to hold and press with hands and feet (and nose), and just enough height to reach surfaces like counters or switches. It’s the kind of design logic also implemented in research platforms such as Toyota’s HSR and the much-beloved-in-labs Stretch robot.
The promise is simple: fetch, tidy, press, and carry. The reality is thornier. Houses are obstacle courses of reflective surfaces, pets, and narrow spaces. Stairs remain a hard stop. Between charges, battery life determines how much work it can accomplish. And its long tail of things that seem so “simple” — opening a pill bottle, unloading a dishwasher, sifting through clutter — pushes perception and dexterity well beyond anything possible with a scripted demo. On the surface, it might seem simple, but if Onero H1 marries hardy hardware to a software pipeline that can learn from user demonstrations and spread skills through updates, it could evolve quickly; without all that, it’s at risk of being an expensive novelty.
Why These WTF Bots Exist in Today’s Consumer Robotics Market
Industrial robotics has been a hot market for years — the International Federation of Robotics reported an all-time high for global installations in 2022 — but the home is still the final frontier, largely conquered only by robot vacuums and mops. This latest crop suggests some kind of next wave, one where mobile manipulation, social presence, and generative AI all combine to take on unstructured, everyday life.
Three takeaways stand out.
- Manipulation is the moat: Compliant arms and five-fingered hands are great, but durability in messy homes is what counts.
- Trust is table stakes: Customers will be looking at safety labels, data practices, and repair options as much as features.
- Business models require definition: What will be included for an upfront price versus subscriptions for software updates, parts, and remote support?
It’s not just the weirdest robots that will win; it’s also the robots that remain useful after the wow factor fades.
Until then, the delightful absurdity is a feature, not a bug. Weird draws crowds. Utility keeps them. These three robots — one that folds, one with a friendship mission, and one that fetches — reveal how close, and how far away, we are from the robot roommate dream.
