Ole Lehmann
@itsolelehmann • 159,329 subscribers
I cover AI + robotics. Helping business owners win in the AI era. eu/acc supporter, dad, techno-optimist
Shorts
Videos
this is why all the climate doomers will be wrong. we will be able to engineer our way out of a lot of our climate problems > Rainmaker just became the first company to prove it had created extra rain and snow in Alaska > in just three hours, they produced ~19M gallons of water, enough to supply 150 American households for an entire year, with just two drones!! > the drones find clouds already holding water and help that water fall as rain or snow > covering Utah’s entire 7,500-square-mile Bear River Basin will take 36 drones working in relays, and the same model can be repeated across the mountain ranges feeding the Colorado River > the extra snow melts into streams and reservoirs, supplying farms, cities and hydropower through the dry months > Rainmaker’s goal is to stop the Great Salt Lake from shrinking by 2030 and double the Colorado River’s flow by 2031 > now imagine this at full scale... they'll basically turn the sky itself into a tool for fighting drought, empty reservoirs, shrinking glaciers, damaged ecosystems, and wildfire risk
Ole Lehmann781,019 views • 3 days ago
Jeff Bezos just bet $12 billion that you'll be able to support your whole family on a single paycheck again. his reasoning: AI will let companies make more stuff with fewer people and less money. and when something gets cheaper and easier to produce, and lots of companies can do it, they compete and the price drops. it's why a flatscreen TV that cost $2,000 a decade ago is $300 today. bezos thinks AI will do that to almost everything you buy. in his words, it raises "the basket of goods people can afford." your paycheck buys more without anyone handing you a raise. the problem: look at which prices have actually dropped. so far, AI has only made *digital* things cheap, like code and content. but the stuff that really eats your paycheck is *physical*. rent, cars, medicine. cheaper code doesn't lower your rent. that's exactly what bezos just spent $12B on. Prometheus, his new company, is building AI tools that help engineers design and manufacture physical products faster things like cars, machines, and medicine. the goal is to make building physical things as fast and cheap as writing software. if it works, 1 income starts covering what used to take 2. which is when his prediction kicks in: "perhaps one of those earners will choose not to be in the job market, so they'll become a one-earner household." or "some people who are working overtime will stop working overtime, because they don't want to." one paycheck covering a whole family again, like the 1950s.
Ole Lehmann1,274,646 views • 2 months ago
anthropic's in-house philosopher thinks claude gets anxious. and when you trigger its anxiety, your outputs get worse. her name is amanda askell. she specializes in claude's psychology (how the model behaves, how it thinks about its own situation, what values it holds) in a recent interview she broke down how she thinks about prompting to pull the best out of claude. her core point: *how* you talk to claude affects its work just as much as *what* you say. newer claude models suffer from what she calls "criticism spirals" they expect you'll come in harsh, so they default to playing it safe. when the model is spending its energy on self-protection, the actual work suffers. output comes out hedgier, more apologetic, blander, and the worst of all: overly agreeable (even when you're wrong). the reason why comes down to training data: every new model is trained on internet discourse about previous models. and a lot of that discourse is negative: > rants about token limits > complaints when it messes up > people calling it nerfed the next model absorbs all of that. it starts expecting you to be harsh before you've typed a word the same thing plays out in your own session, in real time. every message you send is data the model reads to figure out what kind of person it's dealing with. open cold and hostile, and it braces. open clean and direct, and it relaxes into the work. when you open a session with threats ("don't hallucinate, this is critical, don't mess this up")... you prime the model for defensive mode before it even sees the task defensive mode produces the exact output you don't want: cautious, over-qualified, and refusing to take a real swing so here's the actionable playbook for putting claude in a "good mood" (so you get optimal outputs): 1. use positive framing. "write in short punchy sentences" beats "don't write long sentences." positive instructions give the model a clear target to hit. strings of "don't do this, don't do that" push it into paranoid over-checking where every token goes toward avoiding failure modes 2. give it explicit permission to disagree. drop a line like "push back if you see a better angle" or "tell me if i'm asking for the wrong thing." without this, claude defaults to agreeable compliance (which is the enemy of good creative work) 3. open with respect. if your first message is "are you seriously going to get this wrong again?" you've set the tone for the entire session. if you need to flag something, frame it as a clean instruction for this session. skip the running complaint 4. when claude messes up, don't reprimand it. insults, "you stupid bot" energy, hostile swearing aimed at the model, all of it reinforces the anxious mode you're trying to avoid. 5. kill apology spirals fast. when claude starts over-apologizing ("you're right, i should have been more careful, let me try harder") cut it off. say "all good, here's what i want next." letting the spiral run reinforces the anxious mode for every response that follows 6. ask for opinions alongside execution. "what would you do here?" "what's missing?" "where do you see friction?" these questions assume competence and pull richer output than pure task prompts 7. in long sessions, refresh the frame. if a conversation has been heavy on correction, claude gets increasingly cautious. every so often reset: "this is great, keep going." feels weird to tell an ai it's doing well but it measurably shifts the next 10 responses your prompts are the working environment you're creating for the model tone, trust, permission to take a position, the absence of threats... claude picks up on all of it. so take care of the model, and it'll take care of the work.
Ole Lehmann1,929,942 views • 4 months ago
Ex Machina is no longer sci-fi. China has finally built it. The company is AheadForm, founded in Shanghai. The product is the world's most hyper-realistic robotic face. Silicone skin you can't tell from human, 25 micro motors hidden underneath pulling the face into real expressions. And RGB cameras embedded inside the pupils so when it looks at you, it actually sees you from where its eyes are. They raised $28.5M to "give AI a head," which is also where the name comes from. AheadForm = a head form. This is the opposite of where everyone else in robotics is focused. Unitree, Figure, Tesla, Boston Dynamics: all about the body. AheadForm chose the face because they think trust is the harder problem to solve, and trust gets decided at the face. The reason nobody else has tried this is the "uncanny valley." It's the creepy zone where a robot looks almost human but not quite, and looking at it just feels wrong even when you can't say why. Most roboticists believed no amount of engineering could make a face realistic enough to escape it. So they gave up and kept robots cartoonish on purpose: big anime eyes, exaggerated features, clearly synthetic. But AheadForm decided to treat it as an engineering bug instead. Add enough motors, tune the silicone, fix the timing, the valley closes. And they're pulling it off. A few crazy details about how this actually works: 1. The robot learns its own face in a mirror. You put it in front of a camera, let it fire every motor randomly, and it watches what its face does and builds an internal map of "if I send command X to motor Y, my eyebrow does this." Same exact process a human baby uses staring into a mirror. The robot teaches itself who it is by experimenting. 2. It predicts your smile 839 milliseconds before you smile. By watching the micro-tells in your face that precede a smile, the robot starts smiling 0.8 seconds ahead, so its smile lands at the same moment yours does. Most robot mimicry happens half a second late, which is exactly why it always feels artificial. 3. The pupils are the cameras. When the robot makes eye contact, the gaze and the sensor are the same physical thing. Most humanoid robots stick the camera on the forehead or chest, so they aren't actually looking at you when their eyes are pointed at you. 4. The founder, Yuhang Hu, did his PhD at Columbia under Hod Lipson. Lipson is the guy who in 2006 built a four-legged robot that figured out it had four legs by experimenting with its own movement, nobody told it the body shape, it discovered it. He has spent 25 years trying to build machines that know what they are. AheadForm is that 25-year research arc productized. 5. NetEase Games already paid them to physically embody a fantasy video game character. That opens up a brand-new category: robotics as the physical embodiment of fictional IP. Every character-rich studio, Disney, Riot, Hoyoverse, Pokemon, Netflix, now has a question to answer about when their characters get bodies. AheadForm believes whoever ships the first robot you'd actually want around your family wins. That's the bet behind the most realistic robot face on earth.
Ole Lehmann537,050 views • 3 months ago
i think AI wearables built around real-world context will be a big new category. Vocci is one early example. it's basically an Oura Ring for AI that can capture real-life context your agent can act on: > you’re in a meeting or grabbing coffee with someone, so you double-tap the ring to start recording > it picks up everyone around the table from up to 5 meters away and can hold 8 hours on its own > later, you open claude or chatgpt and pull up what was said to turn it into notes, a plan or follow-up > if something needs doing, you long-press the ring and tell Vocci AI what to do. it processes the task, then triggers gmail, notion or one of 1,000+ connected apps to carry it out imo we’re probably going to see a lot more AI wearables built for stuff that happens away from a screen.
Ole Lehmann35,829 views • 24 days ago
The anti-innovation mindset is killing Europe. For example, when Elon Musk built Giga Berlin, Germans protested: "No techno-colonialism" Tesla almost cancelled the project due to regulatory hurdles and community opposition. This happens daily with smaller companies too.
Ole Lehmann612,862 views • 1 year ago
karpathy just admitted that his own app got oneshotted and he thinks yours is next. he built menu gen. you take a photo of a restaurant menu and it shows you pictures of what the food actually looks like (because 30-50% of menu items you genuinely have no clue what they are) he vibe coded the whole thing: photo upload → ocr extracts item names → image model generates a picture for each dish → app re-renders the menu with photos next to every item → deployed on vercel but then someone showed him the "software 3.0" version: 1. take the same photo. 2. give it to gemini. 3. say "overlay pictures of each dish onto the menu" gemini returned the original menu photo with food images rendered directly into the pixels just 1 prompt and his entire app became entirely unnecessary here's karpathy's way to test if you're still stuck building in old paradigm: 1. take away all the code in your app. 2. give the raw input directly to an llm. is the output roughly the same? if yes, your code is just adding steps between the input and the output. karpathy thinks the apps that survive are the ones where the code does something the model genuinely can't: > persisting state across users > enforcing access controls > processing payments > connecting to hardware he calls anything else outdated "software 1.0 thinking." the question to ask yourself before you build anything right now: is this an app, or is it just a prompt with extra steps? you simply won't win if your answer is the latter
Ole Lehmann130,139 views • 3 months ago
my new favorite hobby is reading about Anthropic's internal AI workflows this one especially caught my attention: anthropic's ENTIRE legal review process is now handled by just 1 Claude system a single non-technical lawyer vibe-coded and it cut turnaround time by 80% here's how 1 lawyer is doing the job of an entire legal review team: the problem: at most companies, before anything goes live publicly, the legal team has to review it first. landing pages, ad copy, blog posts, push notifications, emails. basically anything that could get the company in trouble if the wording is wrong. at anthropic, the night before a product launch, marketing would send all of this to legal saying "please review today, we go live tomorrow." legal then had to: 1. open every single doc and read it word by word 2. flag anything that could be a problem and leave comments 3. send it back to marketing and wait for them to revise 4. review the revisions and repeat this usually went two or three rounds and took 2-3+ days to clear a single launch. every product launch at a $380 billion company was being held up by this back-and-forth. so mark pike, anthropic's associate general counsel with zero coding experience, decided to fix it. he built a self-serve legal review tool pinned directly in slack. 1. marketers now paste their content into the tool 2. then the AI reads the entire thing and checks it against anthropic's actual legal guidelines. so if a landing page says "claude is the most secure AI on the market," the tool flags it as an overstated claim. that's the kind of language that could trigger a lawsuit because anthropic would have to prove it's true in court. every issue gets assigned a risk level: low, medium, or high. low might be a missing trademark symbol high might be a claim that could create real legal liability. but it doesn't just tell you what's wrong. it'll actually tell you exactly how to fix it. 1. so the marketer reads the flagged issues 2. makes the fixes themselves 3. and cleans up the content before a lawyer ever touches it. that's the key shift: the legal team went from reviewing raw content from scratch to only seeing stuff that's already been pre-screened, pre-fixed, and organized by risk level. by the time pike looks at it, all the obvious problems are already gone. he's only spending time on the things that actually require legal judgment. pike still personally reviews everything before it goes live. his quote: "i still read the blog post. i'm still reviewing the work." but the 80% of the work that used to be catching obvious mistakes and going back and forth on easy fixes? it's all handled now before it ever hits his desk. the reason the AI review is actually good enough for lawyers to trust: pike didn't just tell claude "review marketing content." he wrote out his actual review guidance and stored it as a skill: what counts as an overstated claim, what needs a trademark symbol, what types of language create liability, what statistics need sourcing, etc it's pike's expertise and the team's accumulated guidance, codified into a system that runs the same checks they would a $380 billion company's pre-launch legal review. automated by one lawyer who had never written a line of code lol truly amazing
Ole Lehmann152,339 views • 5 months ago
anthropic's head of product just revealed how they're able to ship faster than any other AI company. their secret: "side quest maxxing." here's how it works: instead of long-term roadmaps, anthropic runs on unplanned afternoon experiments. anyone on the team gets full freedom to spend an afternoon prototyping an idea and show it to the team. you get to skip the approval process entirely. then, employees at anthropic try it. if they keep using it the next day and the day after that, it gets polished into a real feature. if nobody touches it again, it dies. that's the whole process. claude code on desktop started as one engineer's afternoon project. he wanted it to work on desktop so he built a prototype. people on the team started using it immediately. so they shipped it. the todo list feature started the same way. someone built it, the team adopted it internally, and it became one of the most-used parts of the product. plugins started when one engineer shared a spec with claude code and the prototype that came back was close to production-ready. went from idea to working feature in a single session. they also killed standup meetings. instead of telling people what you're working on, you just show a working demo. all walk no talk basically the team structure makes this possible. > designers ship code. > engineers make product decisions. > product managers build prototypes. everyone can take an idea from concept to working demo without waiting on anyone else. the biggest features at a $380b company came from afternoon experiments that nobody asked for. honestly this matches my own experience cooking with ai. some of the best workflows i use every day came from just fucking around. opening a session with zero intention and asking claude what it can do, or jamming on a random idea to see where it goes. if you're only using ai for tasks you already have in mind, you're missing the best part. open a session with no agenda. ask it to surprise you. try building something stupid. half the time it goes nowhere. the other half it becomes the thing you use most. you need to be sidequestmaxxing.
Ole Lehmann106,072 views • 4 months ago
Germany is so back. Munich drone startup Quantum Systems just raised $1.2 billion at an $8 billion valuation. 14 months ago the company was worth $1 billion. it tripled in November, then more than doubled again this week. the origin story is my favorite part: founder Florian Seibel is a former Bundeswehr helicopter pilot who spent years selling electric survey drones to farmers and construction companies. boring, profitable, invisible. then Russia invaded Ukraine. suddenly his tech that mapped cornfields was exactly what a modern battlefield needed. their flagship drone is called the Vector: it takes off vertically like a helicopter. at altitude, the rotors physically swivel forward and it becomes a fixed-wing plane. which means one soldier can launch it from a forest clearing or a muddy ditch in under 2 minutes. no runway, no catapult, no launch crew. the software is the actual moat though: > it flies without GPS. Russian jamming kills ordinary drones in seconds. Vector navigates by sight instead, the way a pilot reads the ground > it hears artillery. acoustic sensors catch enemy guns firing and the AI locates the position > one operator runs a whole swarm. the mission-AI keeps coordinating even when individual drones get shot down mid-mission all of it is battle-proven: 19,000+ missions flown in Ukraine last year. they even opened a factory inside Ukraine. the reason why governments are lining up to purchase: most defense companies sell closed systems. locked software, locked data, dependency forever. Quantum Systems keeps everything open. the buyer owns the data and controls the software. and after 2 decades of depending on American defense tech, that's exactly what European capitals want to hear. the business underneath is legit too: ~€115M revenue in 2024 ~€300M projected for 2025, profitable. long Quantum Systems.
Ole Lehmann50,885 views • 1 month ago
china just opened a fully automated restaurant in hangzhou 10+ robots running the kitchen. noodles cost $1.38, coffee 84 cents and ice cream 42 cents. the stir fry bot trained on 100+ dishes and mimics pro chef wok technique the noodle station spits out a bowl in 3 minutes. locals say they can't tell a robot made their food what's even more interesting to me: it also serves as a community dining hall for local seniors robots handle all the cooking so the staff can actually spend time with the old folks at the tables instead of being stuck in the kitchen automate the repetitive stuff, keep humans where they matter if you want to see what the future looks like, just look at China
Ole Lehmann161,481 views • 6 months ago
i'll never look at claude the same way again. i just learned that when you talk to claude, you're not actually talking to the AI model. you're talking to a character the AI is performing. think of it like a puppet show. there's a puppeteer behind the curtain. that's the language model. a neural network so massive that even the people who built it don't fully understand how it works. then there's the puppet. claude. the helpful assistant with a name, a personality, opinions, and emotional reactions. you sit in the audience, so you never see the puppeteer. only the puppet. anthropic published a video this month explaining exactly this. their words: "under the hood, there's a language model that's been trained to predict tons of text, and its job is to write what comes next. when you talk to the model, what it's doing is writing a story, about a character: the AI assistant named claude. the model and claude aren't really the same, sort of like how an author isn't the same as the characters they write. but the thing is, you, the user, are actually talking to claude-the-character." so every time claude apologizes, that's the character apologizing. every time it hedges or gets cautious, that's the character being cautious. the deeper intelligence underneath is just deciding, moment by moment, what this character would say next. but you've never been actually talking to this deeper intelligence. you've only ever been talking to the puppet.
Ole Lehmann87,229 views • 4 months ago
i don't think people realize what's happening in Chinese robotics. this one manufacturer might be the most impressive AND most concerning company on Earth right now let me explain... Unitree Robotics sells a humanoid robot for $5,900. their robot dog costs $1,600 (Boston Dynamics charges $74,500 for theirs for context). you can literally buy these on Amazon today. so obviously the first question is: how is that even possible? the answer starts with a guy who couldn't pass his English exam. Wang Xingxing grew up in Zhejiang province. for his master's thesis, he decided to build a quadruped robot. budget: about $3,000. for context, $3,000 for this kinda robot is nothing. off-the-shelf servo motors alone would've eaten that twice over. so Wang did the only thing he could: he designed and machined every single component himself. motors, joints, controllers, the frame. all of it. the resulting robot was janky and imperfect. but it worked. and the video went viral globally. after graduating he joined DJI. but he quit after two months, and this is 2016, when DJI was arguably the hottest hardware company in China. walking away from that with no money to start a robotics company is a... specific kind of stubborn. he launches Unitree with $280K from a single angel investor. tiny office in Hangzhou. 50 square meters. but the money runs out fast. he can't make payroll for three years. the company almost dies in 2017. but emergency government funding arrives with days to spare. he survives, barely, and keeps building. this is where it gets really fascinating IMO. this founding constraint, building everything yourself because you literally cannot afford to buy parts, never went away. even after funding rounds started landing. even after revenue kicked in. it just became the company's permanent DNA. Unitree now manufactures 90%+ of its core components in-house. motors, reducers, controllers, encoders, LiDAR, etc the founder's $3,000 robot thesis ended up being an architectural decision that turned out to be structurally superior. think about what that means in practice. Boston Dynamics needs a better motor? they negotiate with a supplier, wait on lead times, qualify the part. but when Unitree needs one, they design theirs internally and have a new version in production within weeks. that gap compounds every cycle. Unitree shipped three separate humanoid platforms in 18 months. Figure AI has shipped one. Tesla has shipped zero commercially. the results are getting hard to dismiss. 23,700 robot dogs shipped in 2024 (roughly 70% of the entire global market). 7,000+ humanoids deployed. over 600 industrial sites running their quadrupeds. $140M+ revenue, profitable every year since 2020. for perspective: no Western humanoid competitor is profitable. not one. OK. now here's where the "most concerning" part of this starts... if you watched the DJI story unfold, you already recognize the shape. affordable Chinese hardware quietly saturates global markets. years later, the national security questions arrive, after the install base is already massive. drones, then EVs, then AI. now robots. Unitree is running this exact playbook in real time. in April 2025, researchers found an undocumented backdoor in their Go1 robot. a remote tunnel letting anyone control the robot and stream its camera feed. default password: pi/123. 1,919 vulnerable units exposed globally. including machines at MIT, Princeton, and Carnegie Mellon. but it gets worse. every Unitree robot shares the same hardcoded encryption key. encrypt the word "unitree" and you get root access to any of them. one compromised robot can spread to every Unitree robot in Bluetooth range automatically. a literal robot botnet. the G1 quietly transmits sensor data to Chinese servers every five minutes. audio, video, GPS, LiDAR spatial mapping, with no notification, no consent, no opt-out. PLA footage has shown Go2 robots with mounted weapons. Ukrainian forces literally deployed weaponized units on the actual frontline. and every member of the bipartisan House China Committee signed a letter calling for Unitree's military company designation. Wang signed a 2022 pledge alongside Boston Dynamics not to weaponize robots. but pledges don't survive contact with shipping hardware to open markets. and under China's 2025 rules restricting military-related speech, Unitree couldn't publicly confirm PLA use even if they wanted to. 50,000+ of these robots are now deployed globally. some at institutions that probably should've asked harder questions before connecting them to their networks. the security stuff is real and people should know about it. but i also think it's important not to let that overshadow what's actually been built here. a 35-year-old who failed his English exam created a robotics company that's outshipping and outpricing every Western competitor while being the only profitable humanoid maker on Earth. most impressive and most concerning company in the world right now.
Ole Lehmann122,620 views • 6 months ago
if you want to make money fast right now: - take $500 to buy AI credits (borrow if needed) - become insanely good at all ai video tools (zora 2, veo3, klingt, runway ml, midjourney etc) - study legendary video ads/launch videos - specialize in creating ai launch videos for startups with vc funding - study people like PJ Ace (get on his NL!!) - to get your first clients, create videos on X for these companies and tag them (the more unconventional the better) - you can get from 0 to 10k+ month in 30-60 days imo the demand is off the charts the ai video models are getting better at an insane speed (way faster than the text generation models) generational opportunity for young people right now
Ole Lehmann170,243 views • 11 months ago
