El chip del iPhone 18 se calienta menos — y el resto sigue corriendo atrás
La A20 Pro es el chip que no se quema. Samsung y SK Hynix meten $550B en memoria para que no se nos vuelva a dar la RAM. Apple compra un estudio de diseño por el talento, no el logo. Meta entrena a su IA para hablar como gente. Chamath levanta $135M para su taller de código. Esto es lo que se mueve — y por qué le importa a la comunidad.
Apple buys an award-winning design shop for Swift tools — not the logo, the brains
Apple just snapped up an Apple Design Award winner, a small studio whose work centers on Swift tooling. Apple doesn't do this for the press photos; it does it for the code. When Apple buys a shop like this, it's acquiring engineers and the specific knowledge of how they build — the libraries, the frameworks, the quiet patterns that make Swift feel native to Apple platforms.
This is the same play Apple's been running for years: buy the shop, fold the tools into Xcode and the SDK, and let the original team keep shipping. The difference this time is the focus. Swift tooling matters because it shapes how every developer on Apple platforms writes code. A new Swift package manager or an improved compiler flag ripples through the entire ecosystem — the apps, the libraries, the open-source tools that depend on Apple's toolchain.
Why this matters for us: the tools we rely on to build for Apple platforms are quietly being absorbed into Apple's core, and that means the next generation of Swift features will be written by people who actually use them — not by engineers optimizing for quarterly reports.
Samsung y SK Hynix van por $550B para que no se nos vuelva a dar la RAM
Samsung y SK Hynix — los dos gigantes coreanos que mueven el mundo de la memoria— anunciaron un compromiso conjunto de más de $550B para construir nuevas fábricas de chips de memoria. Lo llaman "RAMageddon" y suena a pánico, pero es más bien una apuesta de largo plazo: el mundo se está quedando corto en memoria para entrenar y correr modelos de IA, y las fábricas de hoy no alcanzan.
La memoria no es lo mismo que la CPU, y no la produce la gente de Silicon Valley. Se fabrica en Seúl y Busan, donde los wafers se cortan y se apilan con procesos que nadie fuera de Asia entiende del todo. Lo que cambia ahora es el volumen: a $550B, este es de los compromisos más grandes que se recuerdan en el sector, y posiciona a Corea del Sur como potencia de la IA — no solo en diseño, sino en lo que físicamente se lleva en la mano.
Por qué importa para nosotros: la memoria es lo que hace que los servidores de la gente corran los modelos. Si el precio no baja, los servicios que usamos —la migra app, la plataforma de la abuelita, las herramientas de la comunidad— se vuelven más caros porque los chips cuestan más. Este commit es una señal de que el mercado está intentando resolverlo antes de que nos lo estropee.
Why this matters for us: cuando la memoria se escasea, todo lo que corre en la nube se encarece — y la comunidad que vive con servicios pagados nota el golpe primero.
Generative UI isn't wrong — it's just overhyped.
— evnm.substack.com
#generative-ui-doesn-t-make-sense-676ca9Meta's Brain 2 teaches AI to speak like a person — and gets it
Meta is publishing Brain 2, a model trained to write the way humans talk rather than the way machines sound. The trick is in the data: 300 million pages of human writing — blogs, forums, comments, posts — stripped of the corporate gloss. The model learns to write like a…
La A20 Pro — el chip del iPhone 18 Pro — va más rápido y se calienta menos
Apple está probando un procesador nuevo llamado A20 Pro, y los primeros números sugieren que el iPhone 18 Pro va a sentir la diferencia: más velocidad y menos calor. No es un salto radical, pero es el tipo de mejora que notás en el día a día — apps que abren más rápido, video…
Signadot's new docs: a primer for the non-engineer
Signadot has overhauled its documentation to serve people who ship products but don't live in the terminal. The old docs were written for engineers; the new ones are written for PMs, designers, and founders who need to understand what the platform does before they decide to use it.
The overhaul includes a clearer overview of the core concepts — environments, workloads, and routing — plus a decision flowchart that walks readers through the setup choices without assuming they've read the API reference. There's also a pricing page that breaks down the tiers in plain terms, and a changelog that's actually readable.
This is a smart move for a tool that's trying to cross the chasm from engineering-first to product-first. If you're not the person who writes the YAML, you should be able to open the docs and figure out whether Signadot is worth your time in under five minutes.
Why this matters for us: Signadot is getting easier for the people who run the side businesses and small shops — the ones who need cloud tools but don't have a DevOps team — to use them without a tutorial marathon.
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.
Chamath raises $135M for his AI coding shop and calls CEO
Chamath Palihapitiya founded an AI coding startup and the company just closed a $135M Series A. He's stepping into the CEO chair — so the founder is now the one signing the checks and running the floor.
VCs are still lining up for AI coding tools. This one landed the round on the strength of his name and the product itself. The market hasn't run out of money for the next wave of developer tools, and Chamath is betting it won't for a while.
Why this matters for us: when founders keep writing themselves new rounds, the tools they're building — faster code, faster shipping — tend to be the ones that actually work for the people who use them every day, not the ones that look good in a pitch deck.