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.

good news: AI design-slop is now easy to avoid (if you know what you're doing). you just need to do these 3 things: 1. always give your AI a positive reference AI reaches for the "slop look" when you leave taste up to it. so you need to point it somewhere instead. find a site whose design you love, then get it into your agent one of three ways: > Lafys is a free library of proven website prompts tagged by style and tech stack. grab one that's close to what you want and start from that instead of a blank box > Google Stitch points at any URL and pulls the design system out into a design md. colors, type scale, spacing, layout rules, all written as plain markdown. drop that file in your project and the model builds to that spec instead of guessing one > Figma's Chrome extension copies any live site into fully editable Figma layers. every bit of text, color, spacing, and component comes in as a real layer you can edit. clean it up, then connect the Figma MCP to your agent so it reads the actual design and builds straight from it 2. give it a negative reference too screenshot the quoted tweet and tell your agent to avoid every single thing in it. the gradient, blown-out letter spacing, glassmorphism, "live" badge, etc you're handing it the slop as an anti-target 3. always run design work on extra-high or max thinking in Codex/Claude design is one of those tasks where the extra reasoning shows up directly in what comes back. leave it on default and you get the slop version every time

hilarious how every crypto bro did, in fact, pivot to AI crypto twitter is basically just AI twitter now. which is honestly the most bullish thing i can think of for crypto because every great project i can think of got started in a bear market when nobody cared. > Uniswap shipped in late 2018 when the whole space was a graveyard > Hyperliquid got going right in the middle of the FTX collapse > Ethereum itself got built through the long winter after the 2013 top and you see the exact same thing way outside crypto. Airbnb started in the 2008 crash. Uber right after it. the reason why this happens is pretty simple. you're building right after the price already flushed out everyone who was only there to flip, so most of the downside already happened and basically all the upside is still ahead of you. that's as asymmetric as it gets. the tourists are gone and the dumb money left, but the people who actually care are still here heads down building. and if you're in europe, this is why something like the Solana Summit is worth showing up to right now. Saturday June 13th in Berlin it's a room full of the people who didn't leave. the ones who stick around through the bear winters are usually the ones building the thing everyone's obsessed with two years later.

this could be the breakthrough to making robots 10x more affordable MIT just made an artificial muscle that moves exactly like the real ones in your arm. a soft fiber sits in a charged liquid, a tiny pump runs a charge through it, the charged particles shift, and the fiber squeezes and relaxes. it's the same motion as you flexing your arm, except there's no motor, no gears, and no noise. now, the reason this could be huge for robotic economics: today every joint on a robot moves with its own electric motor and gearbox. a humanoid can have 30 or 40 of them, and they're expensive, heavy, and loud. those motors are one of the biggest reasons a robot costs what it does. and they scale horribly. if you want a joint to push harder, you have to swap in a bigger, more expensive motor, so the price jumps every time you want more strength. the artificial muscle works the opposite way: to make it stronger you just bundle a few more cheap fiber strands together the way real muscle does, so strength goes up while the cost barely moves. interestingly, a similar economic flip happened with computers. early computers filled whole rooms because they ran on thousands of vacuum tubes, which were big, hot, expensive, and always burning out. then the transistor did the same job in something tiny and cheap, and once you could add more of them for almost nothing, computers went from machines only governments could afford to the phone in your pocket. motors could be the vacuum tube of robotics, and this muscle the transistor. if it holds up outside the lab, the most expensive part of a robot gets a lot cheaper which could move robots toward something every small business or household can afford. still lab-stage, but a situation worth monitoring!

if i could long one private company right now, i'd long Unitree. most people don't realize this Chinese company is going to absolutely dominate the robotics industry over the next decade. here's why (inspired by semianalysis's breakdown): first, the robot we're talking about. their flagship humanoid, the g1, is the cheapest humanoid robot in the world by a wide margin. this one fact drives everything else. a year and a half ago the g1 launched at $30-50k. today it starts at $13.5k, a full-size humanoid robot for the price of a used car. and unitree only spends around $9k to build one, so they still make money even at that price. they pull this off for three reasons that stack on each other: 1. they build the most expensive part of the robot themselves. the actuator (the motor inside each joint) is 50-70% of what a humanoid costs. everyone else buys it from a supplier and pays a markup. unitree makes its own, so it pays cost instead. 2. they're in china, sitting on top of the supply chain that already built the world's drones and electric cars. parts are hours away by train, samples arrive the next day, and components run 20-40% cheaper than the west pays. 3. they improve faster than anyone. when they want to upgrade a part, they can have a new version in weeks. a western competitor takes 3 months for the same step, because every handoff is a different supplier in a different country. and cheap is the whole ballgame. a robot only has to clear one bar to be worth buying: it has to cost less than the worker it replaces. it can be mediocre at everything else. the g1 is mediocre right now (it overheats, the arms are weak, it holds 5kg for about 15 minutes), but moving a bin from one spot to another only takes a cheap robot that mostly works. and the g1 just crossed that bar. there are already ~250 of them doing real warehouse jobs, running at under $30 an hour, which undercuts paying a human for the same task. unitree's 10,000th humanoid ships in a few weeks, while tesla still isn't selling optimus to anyone. once a robot costs less than a human worker, a flywheel kicks in: cheap robots sell → the money funds a better next version → the better version handles bigger jobs → that sells even more → which funds the next version. every loop they get cheaper and more capable, and rivals who started more expensive never catch up. we know this flywheel works, because it already built two chinese global heavyweights. BYD owned the battery, the priciest part of an electric car, undercut everyone, and became the biggest EV maker on earth. DJI owned the flight controller, sold cheap, and took 70% of the global drone market. Unitree is running the identical play with robots, and it's only at step one. BYD and DJI both looked this unremarkable right before they took over their industries. Unitree is standing in that same spot today, except the prize this time is bigger than cars or drones. it's human physical labor, the biggest market on earth.