Robots, robots, and more robots
China's racing humanoid robots, Boston Dynamics' bichos for ICE, Nvidia shifting its play, and AI companies warning about cybersecurity — the metal is multiplying. Meanwhile the music labels are fighting Anthropic over copied songs, and the family group chat is still a war zone. Esto te toca.
Sony and Warner sue Anthropic for training Claude on their music
Sony Music and Warner Music have sued Anthropic — the Claude folks — for using their catalogs to train language models. The suit is broad and centers on piracy: the two labels say Anthropic took millions of songs to feed its AI without permission, without paying, without signing anything.
The big labels have always fought for control: royalties, licenses, contracts. What changes now is that the object is no longer an album — it's the sound itself, the frequencies, the timbres — and the user is not a label but an AI startup. The lawsuit puts front and center the question the rest of the world is already debating: who has the right to use others' works to train a model? Until now, the industry defended itself with contracts and technical barriers. Now it needs a court.
The fight isn't just about Anthropic's dollars. If the judge accepts Sony and Warner's logic, the model becomes: any text, image, music, or code trained on protected works without a license is piracy. That changes the cost structure of every AI startup — not just the music ones, the text, video, code, translation, all of them — and forces them to negotiate licenses with the content owners, not just the infrastructure owners.
Why this matters for us: when the big labels win, the cost of building AI gets pushed to the copyright holders — and the people who make the music, the texts, the voices, get to price what gets used. That shifts power away from the tech companies and toward the creators, for better or worse.
Sony y Warner le meten a Anthropic por copiar música
Sony Music Publishing y Warner Chappell presentaron demanda contra Anthropic en el Northern District of California por usar miles de obras con copyright para entrenar Claude. Piden hasta $150,000 por cada obra —y $25,000 extra cada vez que borraron marcas de autoría—, lo que podría llegar a varios miles de millones si el juez los respalda.
Es la tercera batalla grande contra Anthropic. La industria editorial ya le cobró $1.5 mil millones en una arreglo; esta vez son los sellos y los editores de canciones, que dicen que Claude aprendió directamente de sus catálogos. La pelea no es solo por la música: es por qué los grandes estudios pueden permitirse demandar mientras los creadores independientes se quedan sin herramientas de IA y los músicos siguen sin ver un peso.
Why this matters for us: si los sellos ganan, las empresas de IA se blindan con licencias caras y los creadores de la comunidad quedan fuera —pero si pierden, todo lo que usamos para escribir, editar y producir se vuelve más libre.
The tech is real enough to build something that walks and picks up objects. The question is whether it will ever actually make money at scale.
— techcrunch.com
#chinese-car-makers-bet-on-humanoid-robots-cfa8ccNvidia's play is shifting — it's not about raw GPU power anymore
Nvidia is reworking its data center architecture so the network does more of the heavy lifting. Instead of stacking bigger GPUs and hoping the pipes can keep up, the company is putting smarter traffic control inside the switches and fabric that connect them. The idea is simple: move data around efficiently so the chips can actually run.
The shift matters because AI clusters have been bottlenecked by how fast they can talk to each other. More compute means more traffic, and without better routing, the extra silicon sits idle.
Milo Yiannopoulos deported after overstaying his visa
Milo Yiannopoulos was taken by ICE on Thursday and sent back to the UK. The Department of Homeland Security confirmed the removal to Reuters and the Washington Post. He entered legally in 2019, then let his visa expire. When he skipped his immigration hearing, a final removal order was entered on July 22.
Yiannopoulos built his profile as a Breitbart editor, an early Trump booster, and a provocateur whose rants — including bigoted anti-LGBTQ comments — drew constant controversy. He even helped arrange the Trump–Kanye dinner. The deportation ends his US stay; the legal path is clear.
After running $4 billion at a16z, Vijay Pande is betting small
Vijay Pande left a16z's biotech practice last year to start VZVC — a much smaller fund that's AI-native instead of biotech-heavy. The move is a signal: the era of throwing money at everything is over, and the next wave of biology is shifting from discovery to engineering.
Pande says biology has finally moved past the old discovery phase — the shotgun screening, the blind alleys — and is now being built like software. Open datasets, not walled-off ones, are what will actually let AI transform medicine. The old model hoarded data; the new one needs it shared. He also notes that clinical trials remain brutally expensive, a bottleneck that won't go away just because the science is better.
Why this matters for us: the same shift is coming to other health domains — diabetes management, reproductive care, mental health — and the open-data model means better tools for la gente sooner, not locked behind enterprise walls.
AI companies say the cybersecurity apocalypse is months away
Big tech's top researchers have just published a joint paper warning that the next generation of AI tools will make attacks cheaper, faster, and harder to stop. The authors — people at OpenAI, Anthropic, Mistral, and others — are saying the window is "months," not years.
What they're flagging is the obvious thing: AI lowers the barrier to writing malware, phishing, and social engineering. A solo attacker can now do what used to take a team. The paper lays out a few concrete vectors — automated vulnerability hunting, AI-generated lures, and agents that can probe systems without human hand-holding.
DLSS 5 leaks — modders slapping Nvidia's AI upscaler on everything
Code for Nvidia's DLSS 5 has leaked, and the modding community is already running it through games it was never meant for. RenoDX pulled the neural rendering file from an early-access build of NBA 2K27 and got it working in Control, where you can see Jesse Faden's face sharpen up and down depending on the settings. From there it spread to Skyrim, Cyberpunk 2077, and GTA V.
DLSS 5 isn't just a sharper image — it's a neural network that reconstructs frames by guessing what should be there. The modders' tweak lets you dial the AI's influence on things like facial detail, which is why the character's features become more pronounced as you turn it up. The leak shows how the tech actually works: heavy AI inference on the GPU, not a simple filter.
Why this matters for us: if you're on a mid-range card and the mod works, this is free performance — no subscription, no bloatware, just the community doing what they always do.
Quark-gluon plasma forms even in tiny collisions
Scientists at the Large Hadron Collider smashed lead ions together at the very smallest scales and found something unexpected: the quark-gluon plasma forms even in tiny collisions, not just the massive ones. The plasma — the soup of matter that filled the universe microseconds after the Big Bang — doesn't require a big collision to appear. It just needs the energy density.
That flips a working assumption physicists have held for years. The plasma was thought to need a certain minimum size to get hot and dense enough, like needing a big enough pot to bring water to a boil. The new results say otherwise.
How to escape group chat hell without ghosting the familia
Young adults are juggling 20-plus group chats at a time, and the noise is real. The Wire story walks through the tools and tactics for actually taming the chaos—without disappearing from the group entirely.
The practical moves are simple: mute the loud ones, pin the ones that matter, and turn off the pings that steal your attention. You keep the family planning chat open, the cousins' group on low, and the work thread on silent. No big manifesto. Just the basic tools people actually need to live inside a phone that never stops buzzing.
Why this matters for us: our families run on WhatsApp, and if you can't quiet the noise without losing the people, you're stuck. The tools exist—use them so you can hear what actually matters.
China's robot games had some spectacular crashes
The World Humanoid Robot Games wrapped in Beijing this week, and yes — robots fell over a lot. One Honor robot lost a leg mid-sprint. Others tripped and sparked. The five-day event was part of China's push to build humanoid robots that can actually walk and grab things, not just sit on a lab bench.
This is the kind of competition that happens when Beijing throws money at the problem. The US is racing ahead too, but the Chinese side has been betting heavily on humanoids — machines that look and move like people, for factories and warehouses and eventually homes. The crashes are the growing pains.
Hollywood celebs are chasing microdrama cash
A handful of Hollywood names are moving money into microdrama apps — the bite-sized series you scroll through on your phone. Instead of eight-figure checks on massive sets, they're funding short-form episodes that play in vertical format and drop in quick installments. The format is growing fast, and the money is following.
Microdramas are a different beast than a TV show. They're built for the phone, not the couch. Episodes are short, the pacing is tight, and the whole thing is designed to be consumed while you're waiting in line or riding the bus. That's why they work — and why the celebs are showing up. The economics are simpler: lower production cost, faster turnaround, direct distribution.
Why this matters for us:
This format was built for the kind of attention we actually have, and the people making these shows are the ones who know how to keep you watching.
The Z Flip 8 is at its best when there's friction
The hardware is about the same as last year — same hinge, same 3,300 mAh, same 50MP main camera. But the cover screen has a new brain. Samsung finally let the little display run the full Android UI instead of a stripped-down shell. That means the full versions of Google apps, Maps, Wallet, even the camera — all running on the 1.5-inch screen when the phone is folded shut.
For years Samsung held the line on the cover screen: notifications and a few widgets, nothing more. Some fans and Motorola's Razr pushed for the full thing. Now Samsung has given in.
Apple TV ups the price again — $14.99 a month now
Apple is raising the Apple One bundle — which includes Apple TV, Music, Arcade, and Fitness+ — to $14.99 a month, up from $12.99. The increase takes effect September 12.
This is the second hike in two years. The bundle was $9.99 at launch and has been creeping up ever since. The price bump hits everyone who bundles the services together; subscribers who keep Apple TV separate at $9.99 are unaffected.
It's a familiar move for a company that has quietly grown its monthly recurring revenue by increasing prices rather than adding new tiers. The bundle is meant to keep users from drifting to competitors, but the cost is coming off the same wallets — including a lot of working families — every month.
Why this matters for us: when the price of a monthly service creeps up, it eats into the same budget that pays for rent, groceries, and car insurance — and there's no loyalty discount to soften it.
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.
ICE is buying Boston Dynamics robot dogs. Los primos no van a necesitar estos bichos
ICE is shopping for robot dogs from Boston Dynamics — the four-legged machines that trot through warehouses and inspect industrial sites. The agency says they'll improve officer safety, which is the kind of line you hear when you're buying expensive toys. The actual plan is to put these things in detention centers and border operations, where they'll patrol corridors and run sensors instead of people.
Boston Dynamics makes the Spot. They cost around $75,000 each. The agency is planning to spend millions on them. That's the kind of money that could fund a lot of border patrol agents, a lot of translators, a lot of legal aid. Instead it's going to machines that can't read a person's body language and can't tell the difference between a scared kid and a threat.
The bigger story here is the creep of surveillance tech into immigration enforcement. These robots carry cameras, microphones, thermal sensors. They're connected. They send data back to servers. When the feds are buying them, they're not just looking for a new tool — they're building an infrastructure. And the people who end up in their sights are Brown and Black communities.
Why this matters for us: When the government spends millions on robot dogs, it's betting that a machine can do a better job watching our families than a human with a pulse. That's not safety. That's a line in the dirt.