Vibe manufacturers

 

GREG ISENBERG

@gregisenberg

I think we'll see 5x the number of people become hardware and robotics founders over the next 18 months Why? Because you can rent every step, from design to manufacturing: DESIGN - Astra to drive Blender, FreeCAD, and KiCad the way a person would, so you get editable geometry instead of a dead render. Zoo dot dev's text to CAD API if you want it programmatic. - Or have Claude write build123d code, which is Python parametric CAD, and there's a Claude Code plugin that runs it and exports the STL for you. PROTOTYPE - Bambu A1 mini is $299, P1S around $399 on sale. Or skip owning one, JLC3DP prints and mails it for about $20. PCBWay does resin, SLS, and CNC if plastic won't cut it. ELECTRONICS - KiCad for the board, Astra can drive it. JLCPCB fabricates and assembles, often under $100 for a small run. ESP32 for wifi, Raspberry Pi if it needs a brain. ANYTHING THAT MOVES - Unitree sells a Go2 with a full SDK for about $2,500. - Hugging Face put out a $399 open-source biped. - LeRobot gives you the whole train in sim, deploy to real pipeline, and Physical Intelligence open-sourced π0 so you're fine-tuning instead of starting from zero. - Prototype the policy in MuJoCo or Isaac Lab first. MANUFACTURING - Alibaba RFQ for the first 100 units. Check 1688 to see what the factory actually charges domestically, then negotiate. Pietra or Sourcify if you want someone to handle it. FULFILLMENT AND SELLING - ShipBob or Amazon FBA. Shopify for the store, TikTok Shop for distribution, Kickstarter if you want the money before you build it. How to think about starting your own robotics or hardware company: 1. Pick a niche that's already buying weird gear. Cyclists, tabletop gamers, beekeepers, home baristas, dog people with mobility issues. These groups spend money on specific objects and complain in public about what doesn't exist. Or grab ideas off http://Ideabrowser.com 2. Go read the complaints. Reddit, IG etc Search "I wish someone made," "does anyone make," and "modified my." That last one is the best signal, because someone already hacked the product together and you're just manufacturing what they built by hand. Also check Etsy!! If 3 sellers are doing a janky 3D printed version with 400 reviews each, the market is validated. 3. Make one. Describe it to Astra or Claude, get the CAD, print it in ugly gray PLA, use it, fix it. 4. Only go to Alibaba or similar once you've sold a few. Message 10 suppliers through RFQ, take the third cheapest, always pay the $50 for a sample before the real order. 5. Film everything from day 1. The first ugly print, the failed version, the box of 100 arriving. That's your entire marketing budget, and hardware is one of the few categories where people actually want to watch the thing get made. 6. Raise the price. Almost everyone here anchors on what the plastic cost. Your customer is comparing you to nothing, because the alternative is the product doesn't exist. Start at 5x COGS and go up. THIS GOAL OF THIS POST IS JUST HERE TO GET YOUR CREATIVE JUICES FLOWING. Of course, you can build robotics/hardware in a bunch of different ways. One thing I've learned is data couldn't be more important when you're building a hardware/robotics startups. These models learn from first person video of a human doing the task, and that footage doesn't exist for almost any job. So basically you pick one repetitive job people quit over, film someone doing it for 2 weeks, fine-tune π0 on that footage, use organic to sell the first few units and figure out scaling, 5 years ago you needed a factory, a supply chain, and a $1M just to find out if anyone wanted the thing. Now, anyone can become a hardware/robitics founder. And I suspect a lot of people will become one! Vibe manufacturers.

Beatles as engineers

 

Brian Roemmele

@BrianRoemmele

·

1h

Listen, do you want to hear a secret? Do you promise not to tell… The Beatles did not merely write catchy songs. They reverse-engineered the entire 1960s radio machine. They were engineers. Top 40 AM programmers lived by a brutal clock. Every extra second of music was a second stolen from advertising. Songs had to land well under three minutes or they never made the playlist. So the Beatles stripped the traditional intro almost to nothing. No long guitar wander. No polite buildup. "She Loves You," "Help!," and "Can't Buy Me Love" slam straight into the chorus. "A Hard Day's Night" opens with one jagged, instantly recognizable chord that grabs the listener by the collar before the first word is even sung. Change the station and you miss the hook. Most kids did not change the station. Lennon and McCartney also ran a linguistic operation McCartney later called the pronoun strategy. They loaded the lyrics with "I," "me," and "you." Titles were not decoration. "From Me to You." "I Want to Hold Your Hand." "Love Me Do." "She Loves You." The songs were written as a private conversation aimed at one teenager sitting alone in a bedroom or a car. Third-person stories feel like someone else's life. Direct address feels like the band is talking to you. Then they attacked the hardware. Cheap portable transistor radios were flooding teenage rooms. Those radios had two-inch speakers that barely produced bass and drowned in AM static. George Martin and the EMI engineers mixed the early singles accordingly. They crushed the mono masters, cut the low end, and pushed midrange and treble until the close harmonies and ringing guitar lines cut through the tin and the hiss. The records were not mixed for hi-fi living rooms. They were mixed to win on the worst speaker in the house. That combination of instant structure, personal pronouns, and speaker-optimized sound is why those early singles did not just get played. They occupied the dial. And yes for decades it was a secret.

https://x.com/BrianRoemmele/status/2099714451570397400?s=20

Degrees vs knowledge

 

Dustin

@r0ck3t23

A student with a good AI tutor can move through material at a speed a classroom was never designed for, and the credential still takes four years. School was built around fixed constraints. One teacher. One room. One pace. One schedule. One curriculum delivered to everyone at roughly the same speed. AI breaks several of those constraints at once. It can explain the same idea ten different ways. Slow down where you’re struggling. Skip what you already understand. Generate practice on demand. Answer the question you were too embarrassed to ask in class. Give feedback immediately instead of three days later. The student no longer has to move at the speed of the room. If knowledge can be learned faster than the institution can certify it, then the value of the institution shifts from teaching to signaling. A degree increasingly says less about where the knowledge came from and more about the fact that someone survived a long, standardized filter. Employers pay for filters because hiring is uncertain, but filters only survive until a cheaper one predicts performance better. If someone can prove under real evaluation that they can actually do the work, the calendar starts getting much harder to defend. The disruption to education probably does not begin when AI becomes better than every teacher. It begins when four years stops being the cheapest credible way to prove you know what you’re doing.

https://x.com/r0ck3t23/status/2099184734668538112?s=20

My Market Strategy | Stocks Options Dividends

@kfarah9

That's all true but you're missing some major benefits of schools/colleges: 1) The majority of students lack the desire/discipline to self-learn all the way to the finish line. 2) Online learning doesn't provide students with the competition and inspiration of seeing what their classmates can do - and hence what is possible. Of course there is a counter-point: Lots of students go to school/college because they are forced to go - whether by parents or potential employers (pad their resume). Most of those students end up with a degree but without having learned very much. (I've seen this first-hand, both in undergraduate and graduate programs)

Neuralink plus a model

 

Dustin

@r0ck3t23

Neuralink plus a model that can write code on the fly means you could learn a skill in an afternoon. Ten thousand hours on piano. Fifteen years to become a surgeon. A decade to lose an accent. None of that is the knowledge being hard. It exists, written down, sitting there. It takes a lifetime because a human learns through a very narrow pipe. Widen the pipe and the time collapses. Then it gets better. Pair it with a model that builds on demand and you are not limited to skills that already exist. You get ones invented for a problem that showed up an hour ago. Abilities nobody ever developed, because no human lived long enough to be the first. Every skill anyone has ever had was paid for in years. Imagine what people do when that stops being the price.

https://x.com/r0ck3t23/status/2099157147363774634?s=20

Cybercab future

 

Ming

@tslaming

According to the latest ARK Invest's report, the Cybercab isn't just a new vehicle—it’s the master key to unlocking what the firm sees as a multi-trillion-dollar autonomous ride-hailing empire 🔥 The early data out of Austin is wild. Right now, a 15-minute Cybercab ride is running ~40% cheaper than taking a Model Y robotaxi, and undercutting Uber by a massive 50%. Sure, wait times are a bit longer at this stage, but the trajectory is undeniable 👍 Once this hits scale, ARK estimates fares are going to plummet to a mind-blowing $0.25 per mile. We're talking one-third the cost of owning your own car, and a tenth of what you'd pay for a human driver 🤯 But here is the real catalyst: Tesla is going to scale this beast through third parties. Right after the launch, they dropped a “Robotaxi interest form” aimed directly at commercial operators 🤝 Just like ARK mapped out in "Big Ideas 2026", independent fleet owners are going to be the absolute backbone of this value chain, eventually owning and managing the vast majority of the network's vehicles 🆒 The transition isn't coming—it's already here 🔥

https://x.com/davelalande/status/2097439623324561500?s=20

Dave Lalande

@davelalande

I built a planning simulator for anyone thinking about buying a Cybercab and putting it on the network. Not a game about robotaxis. A tool for one question: in my town, at my electricity rate, does one car pay for itself? 179 numbers go into the answer. 17 are published figures. 26 are worked out from published figures. 136 are our estimates — and everyone shows its math and can be changed by you. It's a tool, not a bible. Six towns researched so far, from Dover, Tennessee (1,918 people) to Chicago. The result that surprised me: Dover works — but 84% of its rides are people visiting Fort Donelson National Battlefield. Its residents alone generate 6-8. Close the battlefield, and the town fails, and you can close it in the tool with one edit. Where a published figure exists, we check ourselves against it. Chicago's rides-per-day estimate was within 6% of the number the Sun-Times published, based on city data. Where none exists, the tool says so and names what it searched. It's on my personal lab, http://momatio.com/fleet — free, no signup.