Lo que importa hoy: la gente, no la máquina
Hoy la velocidad de las IAs ya no es el problema — lo que nos separa es saber qué construir. Surgeons, Miami power, Realities, la migra, las bodegas. La tecnología se queda para los que la usan, no los que la inventan. Esto te toca.
Surgeons now steer humanoid robots for their first live-pig operation
A team of surgeons used humanoid robots — the same kind you see on factory floors — to perform a live operation on pigs. The robots were controlled by the surgeons in real time, not by pre-programmed scripts. This is the first time this has been done live.
What makes it interesting is the control method. The robots aren't just mimicking a pre-recorded surgical routine; the surgeon's hands are in the loop, guiding the robot arm the way a surgeon guides a scalpel. This is a meaningful step past the usual demos where a robot does a task once and calls it innovation.
The pig model is old news for surgical research. What's new is using a humanoid robot as the platform — it means the same hardware that walks around a factory could, in principle, walk into an OR and do the same work. The bridge between the two worlds is thinning.
Why this matters for us: when the robots that built our cars start doing our surgeries, the price of both drops — and the people who understand how to run the machines, not just the ones who built them, win.
China's Tianwen-2 has found a quasi-moon — and is about to land on it
The probe has rendezvoused with Kamo'oalewa, a 50-meter rock that loops around Earth as a quasi-moon. It's one of the closest asteroids to us — just 29 times the Earth-Moon distance — and it's been hanging out in our neighborhood for centuries without anyone noticing until now.
Kamo'oalewa's orbit is unusual. It circles the Sun once a year, but wobbles with Earth enough to look like a second moon from the right vantage point. The Tianwen-2 probe — China's second deep-space mission — found it, matched its speed, and parked beside it. The probe sent back the first pictures.
Next step: landing on the surface and collecting samples. That's the part no one's done from a quasi-moon before. If it works, the samples come home to Earth. If it doesn't, you've just spent a lot of fuel chasing a rock that's been right next to us all along.
Why this matters for us: it's a reminder that the cheap parts of space are closer than we think — and the folks who figure out how to grab them first will own the low-hanging fruit for decades.
The bottleneck moved from writing code to knowing what to write.
— magnus919.com
#la-velocidad-de-las-ias-ya-no-es-el-problema-lo-es-saber-que-construir-8a6230Zuckerberg's script is actually working
Mark Zuckerberg just laid out a new playbook for Meta: stop chasing every trend and lean into what Meta is already good at — short-form video, ads, and AI that makes both better. The post breaks down why Meta's recent moves — Threads, Reels, and the AI infrastructure…
Mitchell Hashimoto on why the best AI writes like a person, not a tool
Mitchell Hashimoto — the guy who built Vagrant, Consul, and Terraform — sat down with Alex Alejandre and gave the kind of long-form answer you don't get from a press release. The core insight: the best AI writing sounds like a person, not a machine. The worst sounds like it's…
City Labs nails a first: nuclear power in orbit — Miami-based, privately funded
City Labs, a Miami-based company, has achieved a first for commercial nuclear power in space. The reactor they built is now operating in orbit — the kind of thing that used to take NASA years and billions to pull off.
The company is small. They built the reactor themselves and raised capital through the private markets rather than going public or chasing VC rounds. That path is the one most people actually use: raise quietly, build quietly, launch quietly.
Nuclear in space matters because it gives satellites and stations a power source that doesn't depend on the sun. No more charging cycles. No more shadowed regions going dark. For anything living or operating far from Earth — lunar bases, deep-space probes, orbital factories — it is the right answer.
Why this matters for us: the companies pulling this off are the same kind — lean, private, building real hardware — and they prove that we don't need to be in Silicon Valley to do the kind of work that matters.
Blue Origin raises more — Bezos keeps building the moon base
Blue Origin is raising again, and Bezos is at it: launching rockets, building the Orion 1 habitat for the moon, and quietly turning his old Amazon parking lots into a moon-factory. This is the 150-year plan, not a VC exit play.
The round funds more Starship-class rockets and…
Data is your only moat — and nobody is building for it
OpenAI and Anthropic just published an op-ed in The Atlantic arguing that the real moat in AI is data — not models, not speed, not the latest architecture. The piece is worth reading because it flips a familiar argument on its head: everyone talks about model weight as the competitive edge, but the data layer underneath is what actually compounds.
What makes the piece useful is the contrast it draws between models and data. Models degrade when competitors catch up. Data compounds — every interaction, every correction, every user decision makes the dataset better. A 5-year-old data layer can't be retrofitted by a competitor that didn't exist 5 years ago. That's the kind of asymmetry Brown Forces wins on.
The piece lands on a practical point: companies should think of data as a product, not a byproduct. Treat it like you treat your inventory — track it, invest in it, protect it. Don't let it leak out as a free API call.
Why this matters for us: the moat is the data, the moat is the continuity, and the moat is the discipline to keep it readable across upgrades — exactly what a self-hosted system with multi-decade memory does naturally.
OpenAI ships GPT-5.6 with Sol and Luna on apps and API, plus a new RAG explainer for the rest of us
OpenAI just shipped GPT-5.6 — the headline version lands on apps and the API at the same time. Alongside it, two new models called Sol and Luna joined the lineup, giving you a spread from fast and cheap to heavy-duty reasoning. The whole thing is live today, no beta tag.
…
OpenAI builds a team for the kitchen table — la gente is late to the party
OpenAI is hiring a product manager to shape ChatGPT for families, caregivers, and older adults. The role is new, the posting is fresh — they're building a team, not just a feature. This is the company finally noticing that the kitchen table is where the product lives now.
The bet is real: OpenAI is pushing ChatGPT past the desktop and into the home. A dedicated PM for families means decisions about voice, about privacy, about how the assistant greets a grandmother and how it remembers what the kids said last week. It's the kind of shift that quietly changes the product — the way it speaks, the way it listens, the way it handles a stranger versus someone who's been there for years.
La gente has been talking to chatbots at the kitchen table for a while now. What's different this time is the organizational muscle behind it. A PM for families means someone is accountable for the experience, not just shipping features on a sprint cadence. That's a signal that OpenAI sees households as a market, not a footnote.
Why this matters for us: OpenAI is finally building for the people who actually use these tools — the ones who talk to them while cooking dinner, who ask them for help with the kids, who don't have a degree in tech. If they get the family experience right, the rest of us benefit too.
Anteojos sin cámara: la apuesta de Realities por productividad, no por vigilancia
Realities está lanzando unos lentes inteligentes sin cámara — el primer equipo que apuesta por lo que hace el usuario en la pantalla, no por grabar el mundo. La cámara de los lentes de Meta, WayRay y los demás es la pieza central: captura video, escanea espacios, registra…
Salsa for the bodega: how to make your estimates actually useful
Salvador has been over-estimating for a while now. The pattern is simple: LLMs default to padding, and every small "round up just in case" decision compounds until the numbers stop meaning anything. He's hit it enough times that he now reads "months" as a signal to discount — it's a lower-bound, not a ceiling.
His fix is to write estimates like a person who actually knows what they're doing. When Claude is the engineer and Salvador is steering, things are faster than the default. Days are for things that genuinely take days — real ML training runs, multi-protocol design, sandbox infrastructure — not for anything more than thirty lines of code. If something is multi-week, say so. If unsure, give the low number first and only offer a higher bound when there's a real reason.
The piece is worth a read because it's about the craft of explaining things clearly, not about any specific tool. The rules apply whether you're writing prompts, scoping features, or telling someone how long a project will take. The lesson is the same: stop hedging. Give a concrete number.
Why this matters for us: we're writing for the same people — busy, sharp, tired of the waffle. Brevity with conviction is the voice we're building.
etcd 3.7 is out and it actually fixes the slow writes
etcd is the database that keeps Kubernetes alive — it stores every cluster config, every pod spec, every service. When etcd chokes, the whole cluster chokes. etcd 3.7 is the first major bump in two years and it targets the writes, not the reads. Write latency drops because…
Para la comunidad
Tech affecting the Hispanic community
The stories below land different for our gente — immigration tech, language access, the unbanked, kids of color, gig-worker rights.
La migra: armed and deadly, also kind of a joke
Two stories about ICE landed in the past two weeks, and both tell the same story — just with different camera angles.
In the first, federal agents shot and killed a man and then scrambled to justify the force. The justification was thin. The men with guns made it stick.
In the other, DHS — under former Secretary Kristi Noem — sent house calls to people who'd said mean things about it on social media. A $220 million budget, a cowboy cosplay, and the bureaucracy deciding it's the target of a personal feud.
The contrast is the point. The same agency that can take a life is also the one that can't resist the petty grievance. The heavy hand and the small one are the same hand. Since taking office, the Trump administration has been selling the idea that fighting white supremacy makes you look cool and noble. The budget says otherwise.
Why this matters for us: when la migra is busy with its feuds, it's not watching the fences, the neighborhoods, or the families waiting in court. And when it turns its guns, the excuses come fast.
Una pantalla en la cocina que le dio voz a los chavos
Jaclyn Greenberg bought a Skylight touchscreen calendar — the kind that looks like a picture frame and sticks to the fridge. It turned out to be the hub of her household, not because it's fancy, but because it works without a phone, without an app, without a login. You touch it and something happens.
What's interesting is what it did to her kids. They're touchscreen-native — this is how they've always interacted with the world — and the Skylight gave them a place to put their hands. They can now look at the screen, tap the next soccer game or the dentist appointment, and move it forward themselves. They didn't get a new toy; they got agency. The family calendar stopped being something the parents manage and became something the whole family touches.
That's a small thing. It's also the kind of thing that shows up in houses all over the country, in kitchens with whiteboards and in abuelas' homes with the big wall clock — the same impulse, different materials. The Skylight is just the current version: a piece of glass and metal that tells the house what's next.
Why this matters for us: when tech stops asking the kids to wait for the parents to set things up, the whole family moves faster.
Brown University is in the middle of an AI cheating scandal and the real story isn't the cheating
Brown University is reeling from what it calls the largest AI cheating scandal in its history. The school is investigating whether students used AI tools to ghostwrite assignments — and the scope is so wide it's reshaping how the institution thinks about its own work.
The scandal isn't just that students cheated. It's that they cheated with tools most people now use daily: ChatGPT, Claude, Gemini, and the rest. The problem is the cheating is hard to detect. AI-written work passes the eye test. It reads like a student wrote it. The assignments look correct. The difference is invisible.
Brown is responding the only way it can: by looking at the evidence. The school is examining submission patterns, writing quality, and the telltale markers of AI generation. Some students are being called out. Others are getting the benefit of the doubt. The process is messy and ongoing.
What makes this worth paying attention to is what it says about the tools we use every day. When a prestigious university can't tell what's written by a person and what's written by a machine, the tools have crossed a threshold. They're no longer assistants. They're co-authors.
Why this matters for us: la gente who write reports, grant applications, and policy memos with AI right now — they're probably already in this same gray zone. The rules are still being written, and the institutions haven't caught up.