I always thought humanoid robots look like us because the world is built for humans (doors, stairs, kitchen counters etc) but figure's new release shows another reason, and i think it's the bigger one: > until now, teaching a robot a chore meant collecting data on that exact robot in that exact room, which is slow and expensive and only ever taught it that one thing > figure built an app where people get paid to film themselves doing everyday stuff from their own POV, and about 90,000 people a week are uploading to it now > the robot learned how humans move and handle things from those videos, and then got a bit of extra training on three specific jobs > then they put it in 30 homes it had never seen, with towels, beds and toys it had never seen either, and no training happened in any of those houses > it made the beds, folded the towels and tidied the toys into a basket > the same robot without the human videos finished the job 9% of the time, and with them it finished 56% of the time > and the more human video they feed it, the better it gets at guessing the next move, in a curve you can predict (that's key, it scales!) this only works because the robot has our body. a robot on wheels with a claw can watch you make a bed all day and copy nothing, because it doesn't have the hands or the reach so the human shape also solves the problem robots always had, which is that nobody ever filmed enough robots doing chores. there are 8 billion of us doing them every day though, and now filming us is the training it still fails almost half the time, so nobody is getting a housekeeper this year. but going from 9 to 56 in one step is a big jump the flywheel is flywheeling

something i feel like most people don’t know about, but that i’m very bullish on: world models for robotics. this is how we'll speedrun robotics training. right now, teaching robots requires collecting huge amounts of footage from real machines operating in real environments (which is slow / expensive) but world models are AI video models that can generate simulated versions of that training footage (which is much faster and cheaper) so robots can practice across thousands of environments, failures, and rare edge cases *without* teams having to physically film every example. LTX-2.5 is one example: > generates video and audio of how a scene could unfold > can be adapted to robots, cars, drones, and factories > runs on infrastructure the company controls the endgame is machines that can navigate the physical world better than we do. (think full self-driving, but for every physical task)

this is a marketing masterpiece here's why - taps into the X timeline with built in controversy (@marclou selling his body for sponsors) - calling out customer complains of the main sponsor - positioning his own product as the better solution - while USING his own product to produce the whole thing - the video performing is proof that it actually works to use it for your own marketing it has it all controversy (2x), contrast and proof gg @tibo_maker

it's such a mindfuck to think about how everything you do in codex/claude is just gonna carry over to real life skild is a glimpse of that. you just "prompt" a robot with a video of yourself doing a task, and its general-purpose AI figures out how to do it for you. like imagine running scheduled tasks with it you show a robot how you like your coffee made, then schedule it for 8am every morning. teach it your laundry routine and have it do that every sunday. eventually, even take your dog for its evening walk. i cannot wait to have a recurring task that does my dishes daily lol

The Nebius AI Builder Program is now available. AI isn’t just a model you call anymore. It’s a system you build. And builders, not a few closed labs, will decide what it becomes. The open ecosystem has all the pieces. We want to make it easier to put them together and start building. The program is free, with $400+ in credits and discounts, working code and cookbooks, office hours with engineers, and a community to build with. We’re joined by @NVIDIAAI, @LangChain, @huggingface, @cognition, @OpenHandsDev, @tavilyai, @TolokaAI, @composio, @PrimeIntellect, @MiniMax_AI, @Alibaba_Qwen, and more joining soon. Join with the link in the comments 👇

i almost forgot about the boring company but holy shit has elon been cooking building the tunnel’s concrete walls just became fully autonomous in august! their machine assembles each section in < 1 minute while it keeps digging forward. they've already demonstrated this with nobody inside the tunnel. doing both at once removes the repeated stops that slow tunneling down. so now the target is 1 mile of tunnel/week for under $5 million. and they've just raised $3 billion to keep going. they're already digging in nashville, have dubai under contract, and have been exploring projects in dallas and new orleans. the new UAE agreement alone includes another 150+ km of tunnels. combine this with cybercab and tesla's full self-driving and i feel like elon is on the verge of solving transport. you get dropped off, the car goes to pick up someone else, and underground routes could let more of those trips skip the traffic above. so maybe america can finally do something about all those ugly parking lots lol

AI safety should NOT be a political issue it should be in everyone's interest no matter what political camp you are part of both sides are already cranking up the disinformation and deceptive tactics, it's really concerning to witness

it's starting to leave the screen. my mind is spinning thinking about all the things AI can already do digitally that it’ll soon do physically. image generation → painting on an actual canvas 3D modeling → printing actual objects for you fashion design → sewing actual clothes for you recipe generation → cooking actual meals for you interior design → rearranging your actual home imagine what you could make with a prompt once it can use the tools in your house

humanoids will outnumber us within 15 years exactly because of this btw > elon predicts at least 10 billion humanoid robots by 2040 > getting there means building roughly 2 million every single day for the next 14 years (!!) > that's an enormous problem, the only solution is robots multiplying and building other robots > xpeng just opened the world's first automated production line for advanced humanoid robots > they're literally using robots to build more robots (>80% of the core manufacturing processes are already automated) > xpeng's taking the assembly-line approach they use to mass-produce cars and applying it to humanoids > the goal: produce millions of robots and bring the cost down enough for ordinary people to afford them > xpeng also plans to use those robots' experiences doing real jobs to improve their ai > imagine a robot keeps dropping a cup. engineers can use recordings of those attempts to train its AI to grip the cup better > then send the improved software to other robots running the same system > so one robot's mistakes can help make all the others better > and the more robots you have working in the world, the more examples engineers have to teach them new skills and fix what they're getting wrong so basically robots are giving birth to more robots and they're effectively raising each other, because each robot's mistakes become lessons that make the whole family smarter

I think you could crush it with a travel company organizing no-phone-trips instead of living like a monk in a monastery you would organize an Italy trip, visit old architecture, drink and eat the best local meals, go to parties etc with the strict policy of no phones, no cameras, just 100% being in the moment if you're seen cheating, you're going home (lol) also creates mystery etc because there will be only word of mouth and 0 pics online

i think if we fully simulated a human brain and body like this, it would become conscious. genuinely, why wouldn't it be? imagine we used a more advanced AI model (like GPT-10) to: > recreate the complete anatomy of a human brain, nervous system and body > simulate how it all works, including the electrical signals, chemical activity and feedback between brain and body > give that simulated human a world indistinguishable from ours > let it interact with that world, learn and form memories through the same mechanisms we do what would actually prevent it from experiencing it all as a conscious being? its biology and the world it interacts with would work exactly like ours, after all. and if that human agent became conscious, how would it know its world was simulated? how do we know we aren't just simulated agents ourselves??

everything i'm seeing tells me robotics is really inflecting right now > TONS of talent and capital in silicon valley are pivoting to hardware (because AI has wiped out a lot of software's moat) > robotics + physical AI startups had already raised more than $23B by midyear, versus $27.6B in all of 2025 and only $13.7B in all of 2024 > YC's winter batch funded 35 industrial + defense startups, up from 17 one batch earlier > and this is happening right as foundation models give robots the ability to see, understand and adapt instead of just repeating programmed motions > example: Generalist just showed a robot learning a new task from one physical demo in seconds. basically physical prompting > at the same time, actuators, sensors, cameras and batteries keep getting dramatically cheaper, which is finally showing up in what complete robots cost > a Unitree humanoid now starts at $4,900 and Microduck is taking $399 preorders > this is where the inflection happens: robots become good enough to do useful work AND cheap enough that buying one saves money > jobs that were too difficult or expensive to automate suddenly become worth automating. more businesses have a reason to buy robots > those buyers attract more investment and more companies competing to build better robots for less money > which makes another wave of jobs worth automating, bringing in even more buyers and investment > that's how gradual improvements can turn into rapidly growing demand, with that demand helping fund the next round of improvements > the flywheel is flywheeling also anecdotally, i'm suddenly getting way more robotics sponsorship requests (i got none before) which means robotics companies clearly think normal people are ready to care about robots