Nye's Digital Lab is a weekly scribble on creativity in an age of rapid change.

This week I'm asking about all these back-flipping humanoid robots on my feeds. Do we need them?

So, algorithms of the internet... (as you show me these CES videos of humanoid robotics.)

I ask you...

What problem do these humanoid robots actually solve?

Strip away the stunning demos, the billions in venture capital, Jensen Huang's glittery leather jacket. What specific problem requires a two-legged, human-shaped, $25,000 machine to fix?

I've been down this rabbit hole. Lurked with the LinkedIn Dudes, the sovereign data Bitcoin bros, the Blue Sky Academics arguing machine learning ethics.

The answers the industry keeps repeating don't add up when you actually think about them. I have three responses. One personal, one technical, and one... well, you'll see.

It's the weekend... to the Digital Lab!

Even if humanoid robots could do household chores perfectly, I don't actually want one in my house doing them.

Take laundry. My wife and I would have a better relationship if we didn't have four loads of weekly laundry consuming our lives.

So here's my challenge, robot people! Fix my laundry woes! Can robots make my relationship better by handling the laundry?

Every robotics pitch includes "it can fold your laundry!" Mildly impressive engineering, sure. But that doesn't actually solve the weekly laundry problem.

This needs to be imagined as a system. I want to throw dirty clothes in a hamper and have them magically reappear in my drawer. Clean, pressed, ready to wear. I don't want to be thinking about any intervening steps. In this system, there is not a six-foot humanoid robot carefully grasping each sock and executing a 37-point folding algorithm.

Come on. I don't buy it.

The entire premise is backwards. We've built a world designed for human bodies, and now we're spending astronomical money and engineering effort to create robots that navigate that human-designed world like humans do? But why?

Why not just redesign the systems themselves?

Amazon has over 1 million robots working in its fulfillment centers right now.[^1] Most of them aren't (that's "ARE NOT") humanoid. They're specialized machines.[^2] Invisible infrastructure, optimized for specific tasks. Part of a larger organism operating in concert. Intelligence won't be trapped in a singular robot.

It will be everywhere, in everything.

To me... automation should disappear into the background of daily life, not be a mechanical companion following me around my house.

Having a humanoid robot in your living space means sharing physical space with a large, moving object that makes decisions. Honestly, I don't know how I could have something like that around my sleeping children.

The real solution to household labor isn't humanoid robots, it's better integrated systems. Smart washing machines that know when clothes are dirty. Automated drawer systems. Fractionalized, invisible technology embedded in our environments.

The best automation is the kind you never see, never think about, never bump into at 3am when you have to pee.

OK. Say people actually do want humanoid helpers. (Fine, Nye, you're outvoted. Creepy robots for everyone!)

Can we actually pull it off? The industry's standard response goes:

"Sure, robots aren't perfect yet.

But we just need more data, more computing power, better simulations, camera-based training. The AI boom proved throwing massive computational resources at problems works!

Why should physical AI be different?"

Here's why: The physical world doesn't work like the digital world. The ChatGPT tricks don't automatically transfer to robots that need to actually move around and touch things.

Large language models became amazing because we scraped the entire internet, threw it into massive data centers, and ta-dah!... poetry-writing, code-creating AI. The robotics industry is betting the same "more, more, more" approach will "ignite" embodied intelligence.

I've watched videos of robots learning by watching humans through cameras, practicing in game-like simulations, building 3D virtual environments to test movements.[^4] Simulations especially are going to be huge. (FYI - I write about this stuff all the time, and it genuinely excites me.)

Boston Dynamics' work is genuinely stunning. The things we create now feel like magic.

So here's my pushback! Even the best robots today only hit 98-99% accuracy. In software, that sounds great! On a factory floor? That missing 1% means lost revenue and safety issues. As one analysis pointed out, "If a factory worker was only operating at 99% of their time, they probably would be fired."[^6]

To me: The real problem is range and variability. Ha-Ha! Can't replace the utility players that fast, robot industry!

Training a robot to do one repetitive task is doable. Industrial robots have done that for decades. But the humanoid pitch promises something way more ambitious: robots that work like handymen, recognizing brand-new problems, grabbing the right tool, figuring out what needs fixing, and actually fixing it.

That means handling the full messiness of reality. How do you articulate and learn gravity, lighting changes, moving objects, air pressure, how things feel, how they change over time. This is exponentially harder than the digital worlds where ChatGPT lives.

I'm not a data scientist or a roboticist. There are definitely people who know way more about this than me. But from everything I'm seeing,[^8] this data gap is real. And it's not obvious that cranking up computing power solves the fundamental challenge of getting robots to truly understand and navigate the unpredictable physical world.

So to review: I have a personal problem with humanoid robots in my house. I have technical pushback with "the data moat." Here's where we get to the part that really bothers me.

The industry's ultimate answer to "what problem do humanoid robots solve" is:

TAM—"Total Addressable Market."

And what they mean is humanoid robots are useful anywhere there's a human.

The pitch isn't "robots will help humans do difficult or dangerous work." It's "robots can replace humans in any job a human currently does." The TAM for humanoid robots is, quite literally, the entire human workforce.

So... Let's talk economics.

Even Henry Ford understood that his workers needed to earn enough to buy the cars they were making. That was the basic social contract: automation improves productivity, but the benefits circulate through the economy. Workers earn wages, workers buy products, everyone benefits.

But the TAM model breaks that loop entirely. If humanoid robots replace workers across the board, who's left to buy the products? Who purchases the $20,000-$25,000 household helper robots various companies are promising?[^11] The economic pyramid this creates is mildly concerning.

To me, it's seems what's motivating us isn't "how do we make human life better" or even "how do we solve labor shortages in dangerous industries."

It's "how do we create a massive new market by systematically replacing human workers."

Robotics companies say displaced workers will find new jobs servicing, training, and maintaining robots. Boston Dynamics CEO Robert Playter made exactly this argument, noting that "these robots are not so autonomous that they don't need to be managed. They need to be built. They need to be trained. They need to be serviced."[^13]

But let's do back-of-napkin math... if you need one human to maintain ten robots, and those ten robots each replaced a human worker, you've still eliminated 90% of the jobs.

I tell people angry at things to say it. Put it in your art. Have it motivate your film. So I feel like I have to admit I'm concerned here.

I genuinely believe robotics will play an important role in our future.

But I (think and) hope for innovation at the edges. Smaller, specialized robots for specific applications. Distributed systems, open-source development, peer-to-peer networks of automated devices.

A drone company for neighborhood scanning. Quadruped robots for specific delivery services. Invisible, embedded intelligence in our environments that makes life genuinely easier without the uncanny valley of human-shaped machines.

What I'm skeptical of is the grand vision being sold at CES and in investor decks: that the solution to life's challenges is an army of bipedal robots doing everything humans currently do.

I also can't shake something. Maybe it's a lifetime of science fiction infecting my thoughts.

I think humanoid robots are ultimately being developed so we can be mean to them.

We're building robots we can give hard labor, send into fires, subject to conditions we'd never accept for humans. Let's admit we want servants we don't have to feel guilty about. And if you know the tales of science fiction, that path doesn't usually end well.

So when I see my social media feed clogged with backflipping robots and promises of the "zero labor home," I find myself asking:

What problem are we really trying to solve here?

Maybe I'm wrong. Maybe Atlas and Digit and all their humanoid cousins will usher in an age of abundance and leisure. Maybe three years from now I'll be delighted by my household robot companion, grateful that my wife and I have our Sunday back and no longer have to fold laundry.

But I kind of doubt it. I think I'll still want to throw clothes in a hamper and have them appear in my drawer without any intervening robot steps.

And I'll definitely still want to pee at 3am without an audience.

That's it for this time. If you vibe to the ideas I generate, feel free to subscribe or share. Or if reading books is your thing, I took some of last years' best and jammed it into a printed collection.

Thanks for reading. Make it Happen.

[^1]: Amazon operates 750,000 robots across its fulfillment network as of 2024-2025, with the robot fleet now rivaling its human workforce of approximately 1.5 million. "Amazon hits 1 million robots as AI transforms warehouse operations," Robotics and Automation News, July 2, 2025.

[^2]: Amazon's Vulcan robot, introduced in May 2025, features tactile sensing capability and can handle approximately three-quarters of warehouse items. "Amazon's new warehouse robot has a 'sense of touch' that could see it replace human workers," Live Science, May 21, 2025.

[^4]: Training methods include teleoperation with camera-equipped systems, reinforcement learning in game engines, Gaussian splatting for 3D environments, and simulation-based learning to provide world context. These approaches aim to overcome the data scarcity challenge in physical AI. "Improv 2" transcript.

[^6]: "Improv 2" transcript discussion of robot failure rates and accuracy requirements.

[^7]: The autonomous vehicle industry faces similar data challenges: "The reason why autonomous vehicles aren't all over the road right now, despite from regulation, is the amount of data that they need to be able to collect to run right so you can scrape the entire internet and take all of YouTube and video, but that doesn't give physicality in real world." "Improv 2" transcript.

[^8]: At CES 2026, an Agibot humanoid robot demonstrated by NVIDIA "had trouble standing on the conference center's plush carpets." "Humanoid robots take over CES in Las Vegas as tech industry touts future of AI," CNBC, January 9, 2026.

[^9]: Hyundai announced plans to manufacture 30,000 Atlas humanoid robots annually by 2028 at its Savannah, Georgia facility and deploy them across its factories and warehouses worldwide. "Hyundai plans to deploy thousands of humanoid factory robots," Axios, January 2026.

[^10]: "The same Georgia plant where Hyundai plans to test out Atlas was the site of a federal immigration raid last year that led to the arrests of hundreds of workers, including more than 300 South Korean citizens." "Hyundai and Boston Dynamics unveil humanoid robot Atlas at CES," NBC News, January 5, 2026.

[^11]: Estimated pricing for consumer and commercial humanoid robots ranges widely: Agility Robotics Digit is estimated at $250,000, Unitree H1 at $90,000, and Figure's home humanoid Figure 03 (featured as one of TIME's Best Inventions of 2025) with pricing between $30,000-$150,000. Various sources project consumer models in the $20,000-$25,000 range. "Humanoid robots in 2026: Types, prices, and what's next," Standard Bots; "Innovative Humanoid Robots in 2025–2026," WINS Solutions, December 2025.

[^13]: Boston Dynamics CEO Robert Playter noted that humanoid robots "are not so autonomous that they don't need to be managed. They need to be built. They need to be trained. They need to be serviced." "Boston Dynamics shows Atlas humanoid working at Georgia Hyundai plant," The Robot Report, January 2026.

• Should Robots Look like Us?, November 9. 2025

• My predictions for 2026 (Agents, Open, Edge), January 4, 2026

• Arduino and the Fight for the Edge, December 21, 2025

Nye's Digital Lab is a weekly scribble on creativity in an age of rapid change.

This week I'm asking about all these back-flipping humanoid robots on my feeds. Do we need them?

So, algorithms of the internet... (as you show me these CES videos of humanoid robotics.)

I ask you...

What problem do these humanoid robots actually solve?

Strip away the stunning demos, the billions in venture capital, Jensen Huang's glittery leather jacket. What specific problem requires a two-legged, human-shaped, $25,000 machine to fix?

I've been down this rabbit hole. Lurked with the LinkedIn Dudes, the sovereign data Bitcoin bros, the Blue Sky Academics arguing machine learning ethics.

The answers the industry keeps repeating don't add up when you actually think about them. I have three responses. One personal, one technical, and one... well, you'll see.

It's the weekend... to the Digital Lab!

Even if humanoid robots could do household chores perfectly, I don't actually want one in my house doing them.

Take laundry. My wife and I would have a better relationship if we didn't have four loads of weekly laundry consuming our lives.

So here's my challenge, robot people! Fix my laundry woes! Can robots make my relationship better by handling the laundry?

Every robotics pitch includes "it can fold your laundry!" Mildly impressive engineering, sure. But that doesn't actually solve the weekly laundry problem.

This needs to be imagined as a system. I want to throw dirty clothes in a hamper and have them magically reappear in my drawer. Clean, pressed, ready to wear. I don't want to be thinking about any intervening steps. In this system, there is not a six-foot humanoid robot carefully grasping each sock and executing a 37-point folding algorithm.

Come on. I don't buy it.

The entire premise is backwards. We've built a world designed for human bodies, and now we're spending astronomical money and engineering effort to create robots that navigate that human-designed world like humans do? But why?

Why not just redesign the systems themselves?

Amazon has over 1 million robots working in its fulfillment centers right now.[^1] Most of them aren't (that's "ARE NOT") humanoid. They're specialized machines.[^2] Invisible infrastructure, optimized for specific tasks. Part of a larger organism operating in concert. Intelligence won't be trapped in a singular robot.

It will be everywhere, in everything.

To me... automation should disappear into the background of daily life, not be a mechanical companion following me around my house.

Having a humanoid robot in your living space means sharing physical space with a large, moving object that makes decisions. Honestly, I don't know how I could have something like that around my sleeping children.

The real solution to household labor isn't humanoid robots, it's better integrated systems. Smart washing machines that know when clothes are dirty. Automated drawer systems. Fractionalized, invisible technology embedded in our environments.

The best automation is the kind you never see, never think about, never bump into at 3am when you have to pee.

OK. Say people actually do want humanoid helpers. (Fine, Nye, you're outvoted. Creepy robots for everyone!)

Can we actually pull it off? The industry's standard response goes:

"Sure, robots aren't perfect yet.

But we just need more data, more computing power, better simulations, camera-based training. The AI boom proved throwing massive computational resources at problems works!

Why should physical AI be different?"

Here's why: The physical world doesn't work like the digital world. The ChatGPT tricks don't automatically transfer to robots that need to actually move around and touch things.

Large language models became amazing because we scraped the entire internet, threw it into massive data centers, and ta-dah!... poetry-writing, code-creating AI. The robotics industry is betting the same "more, more, more" approach will "ignite" embodied intelligence.

I've watched videos of robots learning by watching humans through cameras, practicing in game-like simulations, building 3D virtual environments to test movements.[^4] Simulations especially are going to be huge. (FYI - I write about this stuff all the time, and it genuinely excites me.)

Boston Dynamics' work is genuinely stunning. The things we create now feel like magic.

So here's my pushback! Even the best robots today only hit 98-99% accuracy. In software, that sounds great! On a factory floor? That missing 1% means lost revenue and safety issues. As one analysis pointed out, "If a factory worker was only operating at 99% of their time, they probably would be fired."[^6]

To me: The real problem is range and variability. Ha-Ha! Can't replace the utility players that fast, robot industry!

Training a robot to do one repetitive task is doable. Industrial robots have done that for decades. But the humanoid pitch promises something way more ambitious: robots that work like handymen, recognizing brand-new problems, grabbing the right tool, figuring out what needs fixing, and actually fixing it.

That means handling the full messiness of reality. How do you articulate and learn gravity, lighting changes, moving objects, air pressure, how things feel, how they change over time. This is exponentially harder than the digital worlds where ChatGPT lives.

I'm not a data scientist or a roboticist. There are definitely people who know way more about this than me. But from everything I'm seeing,[^8] this data gap is real. And it's not obvious that cranking up computing power solves the fundamental challenge of getting robots to truly understand and navigate the unpredictable physical world.

So to review: I have a personal problem with humanoid robots in my house. I have technical pushback with "the data moat." Here's where we get to the part that really bothers me.

The industry's ultimate answer to "what problem do humanoid robots solve" is:

TAM—"Total Addressable Market."

And what they mean is humanoid robots are useful anywhere there's a human.

The pitch isn't "robots will help humans do difficult or dangerous work." It's "robots can replace humans in any job a human currently does." The TAM for humanoid robots is, quite literally, the entire human workforce.

So... Let's talk economics.

Even Henry Ford understood that his workers needed to earn enough to buy the cars they were making. That was the basic social contract: automation improves productivity, but the benefits circulate through the economy. Workers earn wages, workers buy products, everyone benefits.

But the TAM model breaks that loop entirely. If humanoid robots replace workers across the board, who's left to buy the products? Who purchases the $20,000-$25,000 household helper robots various companies are promising?[^11] The economic pyramid this creates is mildly concerning.

To me, it's seems what's motivating us isn't "how do we make human life better" or even "how do we solve labor shortages in dangerous industries."

It's "how do we create a massive new market by systematically replacing human workers."

Robotics companies say displaced workers will find new jobs servicing, training, and maintaining robots. Boston Dynamics CEO Robert Playter made exactly this argument, noting that "these robots are not so autonomous that they don't need to be managed. They need to be built. They need to be trained. They need to be serviced."[^13]

But let's do back-of-napkin math... if you need one human to maintain ten robots, and those ten robots each replaced a human worker, you've still eliminated 90% of the jobs.

I tell people angry at things to say it. Put it in your art. Have it motivate your film. So I feel like I have to admit I'm concerned here.

I genuinely believe robotics will play an important role in our future.

But I (think and) hope for innovation at the edges. Smaller, specialized robots for specific applications. Distributed systems, open-source development, peer-to-peer networks of automated devices.

A drone company for neighborhood scanning. Quadruped robots for specific delivery services. Invisible, embedded intelligence in our environments that makes life genuinely easier without the uncanny valley of human-shaped machines.

What I'm skeptical of is the grand vision being sold at CES and in investor decks: that the solution to life's challenges is an army of bipedal robots doing everything humans currently do.

I also can't shake something. Maybe it's a lifetime of science fiction infecting my thoughts.

I think humanoid robots are ultimately being developed so we can be mean to them.

We're building robots we can give hard labor, send into fires, subject to conditions we'd never accept for humans. Let's admit we want servants we don't have to feel guilty about. And if you know the tales of science fiction, that path doesn't usually end well.

So when I see my social media feed clogged with backflipping robots and promises of the "zero labor home," I find myself asking:

What problem are we really trying to solve here?

Maybe I'm wrong. Maybe Atlas and Digit and all their humanoid cousins will usher in an age of abundance and leisure. Maybe three years from now I'll be delighted by my household robot companion, grateful that my wife and I have our Sunday back and no longer have to fold laundry.

But I kind of doubt it. I think I'll still want to throw clothes in a hamper and have them appear in my drawer without any intervening robot steps.

And I'll definitely still want to pee at 3am without an audience.

That's it for this time. If you vibe to the ideas I generate, feel free to subscribe or share. Or if reading books is your thing, I took some of last years' best and jammed it into a printed collection.

Thanks for reading. Make it Happen.

[^1]: Amazon operates 750,000 robots across its fulfillment network as of 2024-2025, with the robot fleet now rivaling its human workforce of approximately 1.5 million. "Amazon hits 1 million robots as AI transforms warehouse operations," Robotics and Automation News, July 2, 2025.

[^2]: Amazon's Vulcan robot, introduced in May 2025, features tactile sensing capability and can handle approximately three-quarters of warehouse items. "Amazon's new warehouse robot has a 'sense of touch' that could see it replace human workers," Live Science, May 21, 2025.

[^4]: Training methods include teleoperation with camera-equipped systems, reinforcement learning in game engines, Gaussian splatting for 3D environments, and simulation-based learning to provide world context. These approaches aim to overcome the data scarcity challenge in physical AI. "Improv 2" transcript.

[^6]: "Improv 2" transcript discussion of robot failure rates and accuracy requirements.

[^7]: The autonomous vehicle industry faces similar data challenges: "The reason why autonomous vehicles aren't all over the road right now, despite from regulation, is the amount of data that they need to be able to collect to run right so you can scrape the entire internet and take all of YouTube and video, but that doesn't give physicality in real world." "Improv 2" transcript.

[^8]: At CES 2026, an Agibot humanoid robot demonstrated by NVIDIA "had trouble standing on the conference center's plush carpets." "Humanoid robots take over CES in Las Vegas as tech industry touts future of AI," CNBC, January 9, 2026.

[^9]: Hyundai announced plans to manufacture 30,000 Atlas humanoid robots annually by 2028 at its Savannah, Georgia facility and deploy them across its factories and warehouses worldwide. "Hyundai plans to deploy thousands of humanoid factory robots," Axios, January 2026.

[^10]: "The same Georgia plant where Hyundai plans to test out Atlas was the site of a federal immigration raid last year that led to the arrests of hundreds of workers, including more than 300 South Korean citizens." "Hyundai and Boston Dynamics unveil humanoid robot Atlas at CES," NBC News, January 5, 2026.

[^11]: Estimated pricing for consumer and commercial humanoid robots ranges widely: Agility Robotics Digit is estimated at $250,000, Unitree H1 at $90,000, and Figure's home humanoid Figure 03 (featured as one of TIME's Best Inventions of 2025) with pricing between $30,000-$150,000. Various sources project consumer models in the $20,000-$25,000 range. "Humanoid robots in 2026: Types, prices, and what's next," Standard Bots; "Innovative Humanoid Robots in 2025–2026," WINS Solutions, December 2025.

[^13]: Boston Dynamics CEO Robert Playter noted that humanoid robots "are not so autonomous that they don't need to be managed. They need to be built. They need to be trained. They need to be serviced." "Boston Dynamics shows Atlas humanoid working at Georgia Hyundai plant," The Robot Report, January 2026.

• Should Robots Look like Us?, November 9. 2025

• My predictions for 2026 (Agents, Open, Edge), January 4, 2026

• Arduino and the Fight for the Edge, December 21, 2025