This has been my question for some time. UBI sounds good but where does it come from? Big corporations of course. Where do they get it? From sales of goods of course. Who is there customer? Everyone on UBI of course. ????? Sounds like a version of the innermost loop no one is talking about.
Peter, thanks for the nice write up. Sounds like such a cool trip. Did you see these extended, generalized behaviors at scale in person? I’m finding it hard to separate the real stuff from the hype in this area. The 60-min episode on Boston Dynamics’ Atlas was pretty exciting from the AI training robots angle (I can’t unsee the legion of Atlases learning virtually) … how does what you saw in figure compare to that? Love your work, man.
This reads like a technical breakdown, but it’s really a story about a shift in how civilization builds things!
For most of industrial history, machines were rigid. They did exactly what we told them, no more, no less. Every new task required new code, new rules, new layers of human supervision.
Progress was linear.
Expensive.
Fragile.
What’s emerging now is different.
These aren’t just better robots. They’re learning systems embedded in physical form. Every movement becomes data. Every task becomes training. Every mistake becomes collective memory. Improvement doesn’t stay local as it propagates.
When one machine learns, the whole fleet evolves.
That paradigm changes everything.
Manufacturing stops being a pipeline and starts becoming a nervous system. Factories stop being places and become feedback loops. Hardware starts behaving like software updating, refining, compounding. Vertical integration isn’t about control here. It’s about velocity. The speed at which experience turns into capability.
Data isn’t a byproduct anymore. It’s the core asset.
As recursive production machines building better versions of themselves is how exponential curves enter the physical world.
Most people are still thinking in terms of tools and apps.
This is infrastructure being reborn.
Quietly. Technically. Almost invisibly.
Years from now, people will look back and realize this is where the curve bent.
I actually think sci-fi like Terminator did more good than harm. It forced us to think early about limits, alignment, and responsibility. The real risk isn’t “AI goes rogue.” It’s humans using powerful tools carelessly. Subtlety matters, but so does courage to build wisely and openly.
I'm optimistic too, but with a dose of pragmatic caution.
I don’t think you’re too optimistic about the technology itself, the trajectory feels real, and I’m excited about it. I love the idea of being an old granny lady with a personal robot to help me stay independent - and that’s just one use case.(I'd also like a mech suit that will run me through the forest trails when I'm 95!)
Where I think the conversation needs more depth is on the human-systems side. Much of the optimism around abundance assumes people will smoothly transition once labor is no longer required in the same way, but most people don’t experience the world from a position of optionality. This is incredibly important to understand. Scarcity changes how the brain operates, it narrows bandwidth, compresses decision-making, and makes long-term transitions far harder than silicon valley tech bro culture often assumes. If labor is “freed,” but done in the typical "move fast and break things" way, then we're gonna be breaking economic and psychological things too (livelihoods, identity structures, social stability). If we're not careful, we could move fast and break people.
So, I’m optimistic if we pair this (likely fast) progress if we are super super super super intentional to move fast, but do not break humans. Ai itself can help us with this, but only if we deliberately include perspectives beyond the small slice of the world that already lives in abundance. I have a unique perspective on this because our family got absolutely slammed in the 09 recession, we lost both careers and were 20% above poverty for a bit with 3 small kids. The entire time I was IN scarcity, and also coping doing what I always do, which is studying what was happening, in this case, what it was doing to my brain/perspective. We got out, reinvented careers. This was in part hard work, in part my ability to parse out how my brain was perceiving the world, and what was actually possible. But it's incredibly important to know, the world does not look the same if you're not already in abundance.
This same worry showed up with factories, PCs, and the internet. Jobs didn’t vanish because they evolved. AI will automate tasks, not human ambition. The real risk isn’t unemployment. It’s failing to reskill and retool people for the next economy.
This resonates with what I saw in practice. In my runs, the biggest gains came from tighter task boundaries, explicit verification steps, and tracking failure patterns across reruns. Without that, outputs looked impressive but drifted under load. I now treat failure pattern tracking as a hard requirement before trusting automation in production.
Curious how you keep the result stable across reruns? I would love to compare notes on your exact setup. In my case, this check is what made results repeatable across reruns.
It was so heartening to hear Brett talk about how none of the chinese robot makers are within 2 years of where he is - we can all breath a little easier.
Wild if true, the real headline isn’t humanoid robots exist, it’s reliability + learning at fleet scale: 67 hours autonomous with 1 error, and replacing brittle rule-code with an end-to-end neural controller is exactly how you go from demos to deployment. The choke points now aren’t imagination, they’re safety/liability, data quality, and who gets the productivity dividend. If leasing robots at $300/mo becomes real, we’ll need new social contracts (re-skilling + income mechanisms), otherwise abundance turns into backlash.
Reminds me of The Jetsons cartoons, on one level. Our problems, challenges, opportunities are not technological. Even assuming the pie in the sky time-frames and functionalities are materialized...the reality of anthropogenic global heating is not going to simply disappear. Nation states run by dangerously mentally ill authoritarians with access to weapons of mass destruction won't simply disappear.
The reality that the vast majority of Earth's population is poor won't simply disappear.
Human problems are far more complex and involved to simply be solved by robots and AI.
I've read that many/most young people in America today don't have the attention span to read a book as teachers and professors are starting to simply assume that the students are using AI instead of actually working with the material to acquire their own deep understanding. Obviously, there is still real education going on...especially for the most brilliant self motivated students and teachers...but the trend is for the stupification of the population...I'll stop here before I get into the perverse motivations that determine our political, social, and economic existence in this nation with which we are identified that we call America as if it was itself a planet.
I wish Peter would address the obvious question that must be on everyone's mind but no-one is willing to mention.
In Peter's future there will be the entrepreneurial class which are those who retain agency. The rest will have no agency in a world where AI will do most of the intellectual and physical work. Let's say this entrepreneurial class plus some elites and a few remaining managers and workers make up 10% of the population. The remaining 90% will have nothing to do and will, out of pure boredom, engage in violent, degenerate, and planet destroying behavior.
If you belong to the 90%, what do you think the elites and entrepreneurs will be planning for your future? From their point of view you will be a completely useless and harm causing plob of biomass. A so-called final solution will be engineered for the new useless class.
The problem will be, how do you make the $300 per month if the robots are doing all the jobs?
This has been my question for some time. UBI sounds good but where does it come from? Big corporations of course. Where do they get it? From sales of goods of course. Who is there customer? Everyone on UBI of course. ????? Sounds like a version of the innermost loop no one is talking about.
I was coming here to say this very thing.
There are going to be some major, serious unintended consquences.
Peter, thanks for the nice write up. Sounds like such a cool trip. Did you see these extended, generalized behaviors at scale in person? I’m finding it hard to separate the real stuff from the hype in this area. The 60-min episode on Boston Dynamics’ Atlas was pretty exciting from the AI training robots angle (I can’t unsee the legion of Atlases learning virtually) … how does what you saw in figure compare to that? Love your work, man.
I am confused about if this is a paid advertisement for Figure or a normal post?
How does what they are doing compare to what Tesla is doing with Optimus?
You ignore Tesla's Optimus. It has more data and can accumulate it much faster via their autos.
This reads like a technical breakdown, but it’s really a story about a shift in how civilization builds things!
For most of industrial history, machines were rigid. They did exactly what we told them, no more, no less. Every new task required new code, new rules, new layers of human supervision.
Progress was linear.
Expensive.
Fragile.
What’s emerging now is different.
These aren’t just better robots. They’re learning systems embedded in physical form. Every movement becomes data. Every task becomes training. Every mistake becomes collective memory. Improvement doesn’t stay local as it propagates.
When one machine learns, the whole fleet evolves.
That paradigm changes everything.
Manufacturing stops being a pipeline and starts becoming a nervous system. Factories stop being places and become feedback loops. Hardware starts behaving like software updating, refining, compounding. Vertical integration isn’t about control here. It’s about velocity. The speed at which experience turns into capability.
Data isn’t a byproduct anymore. It’s the core asset.
As recursive production machines building better versions of themselves is how exponential curves enter the physical world.
Most people are still thinking in terms of tools and apps.
This is infrastructure being reborn.
Quietly. Technically. Almost invisibly.
Years from now, people will look back and realize this is where the curve bent.
Maybe we look back and say “Wish we hadn’t made the Terminator movies!” AI viewed and said we must do things a bit more subtly.
I actually think sci-fi like Terminator did more good than harm. It forced us to think early about limits, alignment, and responsibility. The real risk isn’t “AI goes rogue.” It’s humans using powerful tools carelessly. Subtlety matters, but so does courage to build wisely and openly.
I'm optimistic too, but with a dose of pragmatic caution.
I don’t think you’re too optimistic about the technology itself, the trajectory feels real, and I’m excited about it. I love the idea of being an old granny lady with a personal robot to help me stay independent - and that’s just one use case.(I'd also like a mech suit that will run me through the forest trails when I'm 95!)
Where I think the conversation needs more depth is on the human-systems side. Much of the optimism around abundance assumes people will smoothly transition once labor is no longer required in the same way, but most people don’t experience the world from a position of optionality. This is incredibly important to understand. Scarcity changes how the brain operates, it narrows bandwidth, compresses decision-making, and makes long-term transitions far harder than silicon valley tech bro culture often assumes. If labor is “freed,” but done in the typical "move fast and break things" way, then we're gonna be breaking economic and psychological things too (livelihoods, identity structures, social stability). If we're not careful, we could move fast and break people.
So, I’m optimistic if we pair this (likely fast) progress if we are super super super super intentional to move fast, but do not break humans. Ai itself can help us with this, but only if we deliberately include perspectives beyond the small slice of the world that already lives in abundance. I have a unique perspective on this because our family got absolutely slammed in the 09 recession, we lost both careers and were 20% above poverty for a bit with 3 small kids. The entire time I was IN scarcity, and also coping doing what I always do, which is studying what was happening, in this case, what it was doing to my brain/perspective. We got out, reinvented careers. This was in part hard work, in part my ability to parse out how my brain was perceiving the world, and what was actually possible. But it's incredibly important to know, the world does not look the same if you're not already in abundance.
Why no mention of Tesla and Optimus? They are surely not 10 years behind.
I'm waiting for the robot that can muck out my animal pens and stack hay bales on the farm. You know, the back breaking, hard work!
"You come home from work and your robot has already meal-prepped dinner". But will there be any jobs left? Most people will be unemployed by then.
This same worry showed up with factories, PCs, and the internet. Jobs didn’t vanish because they evolved. AI will automate tasks, not human ambition. The real risk isn’t unemployment. It’s failing to reskill and retool people for the next economy.
This resonates with what I saw in practice. In my runs, the biggest gains came from tighter task boundaries, explicit verification steps, and tracking failure patterns across reruns. Without that, outputs looked impressive but drifted under load. I now treat failure pattern tracking as a hard requirement before trusting automation in production.
Curious how you keep the result stable across reruns? I would love to compare notes on your exact setup. In my case, this check is what made results repeatable across reruns.
It was so heartening to hear Brett talk about how none of the chinese robot makers are within 2 years of where he is - we can all breath a little easier.
Wild if true, the real headline isn’t humanoid robots exist, it’s reliability + learning at fleet scale: 67 hours autonomous with 1 error, and replacing brittle rule-code with an end-to-end neural controller is exactly how you go from demos to deployment. The choke points now aren’t imagination, they’re safety/liability, data quality, and who gets the productivity dividend. If leasing robots at $300/mo becomes real, we’ll need new social contracts (re-skilling + income mechanisms), otherwise abundance turns into backlash.
Peter,, if you have Skippy, are you changing your name to Joe?
Reminds me of The Jetsons cartoons, on one level. Our problems, challenges, opportunities are not technological. Even assuming the pie in the sky time-frames and functionalities are materialized...the reality of anthropogenic global heating is not going to simply disappear. Nation states run by dangerously mentally ill authoritarians with access to weapons of mass destruction won't simply disappear.
The reality that the vast majority of Earth's population is poor won't simply disappear.
Human problems are far more complex and involved to simply be solved by robots and AI.
I've read that many/most young people in America today don't have the attention span to read a book as teachers and professors are starting to simply assume that the students are using AI instead of actually working with the material to acquire their own deep understanding. Obviously, there is still real education going on...especially for the most brilliant self motivated students and teachers...but the trend is for the stupification of the population...I'll stop here before I get into the perverse motivations that determine our political, social, and economic existence in this nation with which we are identified that we call America as if it was itself a planet.
Thanks to anyone who has read this far. ; )
Thanks for sharing. Fantastic stuff. Cannot wait
Thank you!
I wish Peter would address the obvious question that must be on everyone's mind but no-one is willing to mention.
In Peter's future there will be the entrepreneurial class which are those who retain agency. The rest will have no agency in a world where AI will do most of the intellectual and physical work. Let's say this entrepreneurial class plus some elites and a few remaining managers and workers make up 10% of the population. The remaining 90% will have nothing to do and will, out of pure boredom, engage in violent, degenerate, and planet destroying behavior.
If you belong to the 90%, what do you think the elites and entrepreneurs will be planning for your future? From their point of view you will be a completely useless and harm causing plob of biomass. A so-called final solution will be engineered for the new useless class.