Robot Update: Rosey’s Not Here Yet

But Figure AI just gave us a glimpse of someday.

Source: The Jetsons Episode 1

Since the 1960s, Rosie the Robot from The Jetsons represented a distant future with a humanoid assistant gliding through domestic chaos, emptying dishwashers and keeping the home running effortlessly.

Rosey came with a Brooklyn attitude and a charming personality but represented a future that seemed a long way off.

But a few weeks ago, Figure AI released another of its impressive humanoid demonstrations with its latest Figure 3 and its brains, Helix 02. Yes - we’ve seen awesome video snippets in the past from Figure like making a Keurig coffee dating all the way back to ancient times (2024).

So why is this one different?

Because the latest Figure AI demonstration is not that a robot moved dishes.

It emptied it. It reloaded it. It closed the door. It started the cycle.

This Figure robot operated inside an unstructured, semi-chaotic environment and completed a multi-step task without visible resets or human guidance.

And that is the transition that makes Rosey possible.

[embed]Source: Figure (YouTube)

Why the Dishwasher is Important

Aside from the fact that loading a dishwasher is in my top three use cases for my personal robot, it is not a factory task.

It is not standardized. Dishwashers get loaded inconsistently and often not efficiently. The numbers of plates, bowls or glasses vary from meal to meal. The same for the utensils and pots and pans - nothing is uniform and from a robotics standpoint that variability is the entire problem.

Industrial automation has been highly successful in environments defined by precision and repetition. Automotive welding, semiconductor fabrication, pallet transfer — these tasks operate within tightly controlled parameters.

Domestic manipulation is different. It requires perception under clutter, object recognition across arbitrary configurations, adaptive grasping for fragile materials, and spatial reasoning in three dimensions.

More importantly, it requires sequencing — determining not just how to move an object, but which object to move first, where it belongs, and how each action affects the next.

This is not something that can be effectively coded across infinite possibilities.

This demonstration shows that Figure’s Helix 02 provides inference that is able to operate in the unpredictable geometry of humans.

And the prospect of having my personal Rosey seems closer than ever.

How Fast Figure Has Actually Moved

I’ve been waiting a long time for so many innovations — space travel, robots, longevity. The Star Trek world with all of the great technological advancements was always just in the future.

But over the past couple years, the speed toward the future has exploded forward, largely due to the advent of artificial intelligence.

Figure AI was only founded in 2022 with its first appearance of Figure 01 in 2023. It started with basic walking and programmed structured task execution. Movements were mechanical and planned with most of the intelligence stack leaning heavily on external large language models like OpenAI.

That coffee demo of Figure 01? January 2024.

[embed]Source: Figure (YouTube)

But then Figure transitioned away from relying on OpenAI integrations and began building its own vertically integrated control architecture — Helix. Helix 01 started that shift and Helix 02 expanded it with extended learned control across locomotion, balance, and manipulation rather than isolating those systems into separate code stacks.

At the same time, the hardware evolved. Figure 02 improved actuator density and reliability. Figure 03 refined gait stability and — more importantly — dexterity.

The hands are where the progress becomes visible. The newer hand design shows increased articulation and finer finger control, allowing adaptive handling of fragile objects like glassware.

Figure 03 integrates the hardware and intelligence together making that dishwasher demo possible. And all of this happened less than 24 months from the Figure 01 coffee demo.

The Factory Floor

Figure AI hasn’t just been working on loading the dishwasher over the past few years. They have actively deployed robots to factory floors.

In many ways the factory deployment is an easier starting point.

Factories provide controlled but economically meaningful environments. Tasks are repetitive enough to measure performance yet varied enough to stress-test adaptability. Reliability can be measured and corrected in real use cases.

Figure’s most notable deployment has been with BMW at its Spartanburg plant. There, humanoid units have operated across full shifts handling logistics and part movement within live production workflows. These are not choreographed lab exercises but rather they are factory hours working.

Beyond BMW, others have also tested but numbers are relatively small. However, the goal is not viral scale but operational learning.

[embed]Source: Figure (YouTube)

With the progress of Figure 03, capabilities across factory floors expand and more real-world deployment is expected in 2026. Brett Adcock, Figure’s CEO, is close lipped on actual numbers, but has teased in the thousands for this year and 100,000 by 2029.

The Tipping Point: When Does This Become Real?

The practical question behind every robotics demonstration is simple: when does this move from impressive to useful?

A robot emptying and reloading a dishwasher is compelling, but it does not automatically signal consumer readiness. The tipping point is not the first successful execution of a task. It is the moment when performance, cost, and reliability converge to make automation rational.

In industrial settings, that threshold is easier to reach. Labor costs are defined and many tasks are structured. Variability exists, but typically within measured constraints. A humanoid robot does not need to be flawless to justify deployment on a factory floor.

Incremental reliability improvements can justify scaling within these controlled environments.

Domestic adoption operates under a different standard. A home is a dynamic system constantly shifting — people are unpredictable. The tolerance for failure is low and so a consumer robot must reach a higher reliability threshold before it becomes broadly viable.

Perfection is not required.

Technology adoption rarely waits for flawless performance. It accelerates once systems become consistently competent and visibly improving. Humanoid robotics is likely to follow this pattern and early deployments will not be comprehensive household assistants.

They will be narrow, task-oriented systems that improve over time through software updates.

As the hardware scales, costs decline and as operating hours accumulate skills and reliability constantly improve. The tipping point is not a singular breakthrough moment but a gradual convergence of these forces.

But with artificial intelligence, things are not necessarily limited to real-world trial and error. Virtual physics models create near infinite opportunity to pre-train. And it is this component that begins to create a compressed iteration cycle of improvement.

This is part of the reason that Figure AI has made such tremendous strides in the span of just two years.

Who knows where the next two years gets them?

Figure AI is Not Alone

If Figure represents one side of the humanoid acceleration curve, Tesla represents the other.

In early 2026, Tesla announced the discontinuation of the Model S and Model X. On their latest earnings call they explicitly indicated the purpose was for production capacity and engineering focus for Optimus.

That is not incremental positioning. It is strategic commitment.

Tesla has already begun deploying Optimus internally within its own factories. Early use cases center on logistics and repetitive material handling follows the same model as Figure’s factory deployments — controlled spaces where iteration can occur under real economic conditions.

According to Elon Musk, Tesla is hoping to deploy a legion (10,000) of Optimi (his preferred plural is Optimi). Over the next few years, with rapid advancement, he expects production to soar into the millions.

Tesla has infrequently released Optimus demonstrations and almost none of Optimus Gen 3. Musk cites competitive pressures for the secrecy and he’s probably right. At this year’s CES, 38 companies featured humanoid robotics categories.

Don’t forget Boston Dynamics

For decades, Boston Dynamics represented the gold standard of robotic locomotion. Atlas demonstrated extraordinary balance, mobility, and dynamic control long before humanoid robotics became commercially fashionable.

Yet Boston Dynamics historically relied on highly engineered control stacks rather than large-scale learned models. But as the competition quickly demonstrated, integrated learning-based models are the real path to task execution, Boston Dynamics has shifted to the AI-driven control architectures moving beyond scripted routines.

Their Atlas humanoid seems ready for its factory floor trials after showing several demos.

Their recent majority acquisition by manufacturing powerhouse Hyundai is also a clear indicator that Boston Dynamics/Hyundai are headed to production.

[embed]Source: Boston Dynamics (YouTube)

But Then — There’s The Rest of the World

The humanoid robotics race is not confined to Silicon Valley and Boston.

China has over 30 prominent companies and currently leads in unit volume and cost compression. Companies such as Unitree, UBTECH, Fourier Intelligence, and others have accelerated hardware development and introduced lower-cost humanoid platforms. at a pace that would have seemed unrealistic just a few years ago.

In 2025, Chinese companies delivered over 15,000 humanoid robots.

Chinese robots have also dominated viral visibility. Rapid walking speeds, backflips, coordinated movement routines, and dynamic balance demonstrations regularly circulate online. The mechanical progress is undeniable. Actuator performance, torque density, and control precision have improved significantly.

[embed]Source: Unitree (YouTube)

However, it is important to distinguish between choreographed execution and adaptive autonomy.

Many of the most widely viewed demonstrations rely on pre-programmed motion sequences or operate in carefully structured environments. The robot may perform impressive dynamic movements, but the task context is often constrained.

High-level reasoning — interpreting a novel scene, planning a new sequence of actions, adapting to unexpected variation — remains a more difficult hurdle.

That does not mean Chinese firms are behind. It means the competition is asymmetric.

China’s advantages lie in manufacturing scale, cost reduction, and rapid hardware iteration. Western firms, particularly in the United States, have concentrated more heavily on integrating large-scale AI models into embodied systems. Over time, those two vectors may converge.

The Direction Is Clear

The most serious contenders in humanoid robotics are no longer focused on mechanics alone.

Figure and Tesla have both landed on production ready forms with many levels of articulation. And in those forms, dexterity and movements are capable of fine manipulation. But each has paired that physical precision with intelligence and reasoning systems that can interpret disorder.

Boston Dynamics, long the benchmark for mechanical excellence, has recognized this shift and joined the movement toward model-driven control. Its early lead in locomotion remains formidable, but the competitive frontier has moved upward into software.

Chinese manufacturers bring something equally powerful to the table: manufacturing scale, cost compression, and state-supported development cycles that accelerate hardware iteration. That advantage should not be underestimated. When volume scales, learning accelerates.

Rosey is imminent

The direction is clear.

Capital is committed. Production lines are being retooled. Factory deployments are increasing. Skill libraries are expanding. And unlike previous robotics cycles, this one is fueled by AI systems that improve continuously rather than statically.

We have already seen how quickly software-defined systems can compound in the physical world. Tesla’s Full Self-Driving improved not through isolated breakthroughs but through relentless iteration, data accumulation, and deployment at scale. Humanoid robotics appears to be entering that same feedback loop.

Each of these contenders will hit that ‘long tail’ list of problems to contend with like those unexpected spills or lost items or kids and pets. This last 10% of reliability can take longer than the first 90%, but it doesn’t have to be perfect. It must be safe - sure, but it’s also not traveling at 70 miles per hour.

Taken together, we appear to be at a major convergence point of mechanical capabilities and artificial intelligence. With both, the humanoid is likely impossible.

Together, humanoid robots are inevitable.

Here is my subjective estimation of when full production and mass deployment of humanoid robotics happens for the domestic market:

• 2 Years — 10%
• 3 Years - 25%
• 4 Years — 50%
• 5 Years - 75%
• 10 Years - 99.9%

While the exact timeline is uncertain, the destination is not. Tesla and Figure appear to have clear leads, but the others are in a full sprint too.

So, Rosey is not quite here yet, but you may want to make room in your home for her soon.

Want more? Here are some related articles:

**The 2026 CES Autonomous Vehicle Explosion: Is Tesla Cooked?**

**Are the Magnificent 7 Dead in 2026?**

**Tesla Stock Refuses to Crash — Because the Market Finally Sees What’s Coming**

**If You Think NVIDIA’s Run Is Over, You’re Not Watching Where the Industry Is Going**

**Google’s Stock Won’t Stop Climbing — And Why It Might Just Be Getting Started**

Follow me on Substack or online at evervests.com.

Images Source: Unless specifically noted, all images were created by Evervests.com by AI

Disclaimer: This content is for informational and educational purposes only and should not be considered financial, investment, or trading advice. The views expressed are based on publicly available information and personal opinion at the time of writing. Markets and conditions may change. Always perform your own research, verify data independently, and consult with a licensed financial advisor or investment professional before making investment decisions. The author may hold positions in the securities or assets discussed.

A message from our Founder

Hey, Sunil here. I wanted to take a moment to thank you for reading until the end and for being a part of this community. Did you know that our team run these publications as a volunteer effort to over 3.5m monthly readers? We don’t receive any funding, we do this to support the community.

If you want to show some love, please take a moment to follow me on LinkedIn, TikTok, **Instagram. You can also subscribe to our [weekly newsletter](https://newsletter.plainenglish.io/). And before you go, don’t forget to clap and follow** the writer️!