Good morning. Somewhere between “please take off your shoes” and “don’t move the antique furniture,” Airbnb hosts may now need a new rule: no giant robots in the living room. A San Francisco startup raised $300 million to build bots that do housework, and needed a realistic lab to test them. Naturally, they booked a rental. What followed was chaos. The angry host says more than 30 people came and went, furniture was smashed, cables crawled everywhere, and a 6 foot Roomba on treads led the assault. Total damage: $12,000, which is still cheap for an R&D lab.
Also, Anthropic’s Claude Fable 5, a “safe” version of Mythos, is now open to the public.
Age reversal is moving from hype to human trials. David Sinclair plans to test an oral reprogramming drug in volunteers as part of the $101 million XPRIZE Healthspan Competition. The prize goes to teams that can show a 10 year improvement in immune function, cognition, and muscle performance after a year of treatment. Sinclair says his drug aims to push age restoration across the whole body, not just one organ. Critics are skeptical. Sinclair hasn’t published animal data or disclosed what’s in the drug, and some scientists say he oversells age reversal. Others have tried chemical reprogramming in animals and hit toxicity issues. Everyone wants a pill that makes them younger, which is exactly why proof matters before anyone starts selling time in a bottle. (Xprize, MIT Tech Review)
AI actors now need warning labels. New York just made it illegal to use AI generated people in ads without clearly saying so. Brands can still use fake people to sell real products, but the new law requires them to be labeled as “synthetic performers” to disclose they didn’t have a mother. A first violation costs $1,000, and repeat violations jump to $5,000. SAG-AFTRA backed the law because actors saw this coming from space. Advertisers fought it because telling people your spokesperson is fake does tend to kill the vibe. (AP)
Quid is an AI-native consumer and market intelligence platform built to fit inside the tools your enterprise already runs on.
No new tab addiction. No separate login graveyard. Billions of data points from top social channels, markets, and patents, delivered where decisions actually get made.
👀 closer look
Cheap AI models are starting to look good enough. The AI boom ran on the idea that bigger models are better and whoever burns the most tokens wins. Now the bills are arriving, and companies are struggling to prove the spend did anything useful. Some insiders believe 80% of AI workloads could shift to models that are 99% cheaper within the next 12 to 18 months, provided there’s any hardware available. Harvey reportedly cut inference costs 3x without hurting quality by sending harder legal work to Claude Opus and easier tasks to cheaper models. That could reshape big labs if customers stop paying premium prices for tasks smaller models can already handle. If only there were some kind of AI model router that could look at each prompt and send it to the cheapest model smart enough to do the job. (Silicon Angle, TechCrunch, Openrouter)
China is connecting its data centers into one giant AI brain. Beijing is prepping a $295 billion plan to link scattered computing hubs across the country, turning isolated data centers into a national AI network. State firms like China Mobile and China Telecom would run most of it. The plan calls for at least 80% Chinese tech, giving Huawei a front row seat while leaving Nvidia outside checking if its badge still works. The network is expected to come online by 2028, giving companies broader access to high performance computing. But the real ambition goes beyond data centers. China may tie the network directly into the power grid, pushing total investment to at least 5 trillion yuan. This is what the AI race looks like when a country stops buying shovels and starts building the railroad. (Bloomberg)
👽 Welcome to the Future
The next presidential election will be fought over AI. AI companies are creating so much wealth that Washington is starting to ask who gets to keep it. Most of that money flows to companies and shareholders, while workers get a much smaller slice of the economy. That’s how AI moves from the boardroom to the ballot box. Democrats, Republicans, OpenAI, and Anthropic are all circling the same idea: the public might need a stake in the AI economy. The fight is whether Big Tech pays people later through taxes or gives them ownership before the money disappears upstairs. Silicon Valley promised AI would change everything. Voters are asking if “everything” includes their bank accounts. (Noema)
MIT built a bracelet so robots can finally get a grip. Humanoids can put on an awesome show right up until someone asks them to use their hands. Grasping a cup or a scalpel is still hard, which makes housework and surgery seem like complete moonshots. MIT’s answer is to put an ultrasound wristband on a person, then let a robotic hand copy what happens under the skin. The system watches muscles, tendons, and ligaments as they move, then turns that into motion the machine can mimic. Creepy, right? In tests with eight volunteers, it copied all 26 letters in American Sign Language within 120 milliseconds. The bigger play is building giant datasets of human hand motion so robots can learn dexterity without someone guiding every move. We built robots to escape chores, now we’re stuck teaching them. (Fortune)
fun stats
💩 30%. Developers who knowingly ship vulnerable AI code to meet deadlines. Meanwhile, 70% say AI generated code is less secure.
🦾 $200 million. What Standard Bots raised to scale AI native industrial robots in the US. Last year China installed 9x more industrial bots than America.
🚀 $1 trillion. Valuation OpenAI is reportedly targeting for its IPO. SpaceX wants $1.75 trillion. Silicon Valley has moved beyond billions.