Pablo Sanchez

🔴5 farklı yapay zeka ajanı 145 m2’lik bir dairenin tüm mekanik ve elektrik tesisatını tasarladı Her agent farklı bir sistem üzerinde çalıştı: temiz su, pis su, havalandırma, elektrik ve yangın tesisatı Toplam 4871 farklı tasarım denendi. Sistemler birleştirildiğinde 11 tesisat çakışması tespit edildi. 6. yapay zeka ajanı devreye girerek tamamını giderdi. İlk tasarımda 139 saniye olan sıcak su bekleme süresi ise 6 saniyeye düşürüldü. Mause imleçinin ne kadar hızlı çalıştığını görebiliyor musunuz?

We will all be learning in coming weeks & months, that most human-made software: ▫️ Was really slow ▫️ Had many security vulnerabilities ▫️ Was under-specified and incorrect AI will rewrite it all. Many times. It will make it fast, cost-efficient, and secure. It will think about failure modes we couldn't wrap our heads around. It will prove, formally verify, load test, harden. It will turn memory-unsafe foundations into robust, optimized ones. If you ask, it'll thoroughly document it, teach you about it, make diagrams, animations, and slide decks. In many ways, we are witnessing the true birth of software engineering.

Introducing Claude Fable 5.2 and Claude Mythos 5.2, designed to tackle the most daunting technical challenges. They’re capable collaborators on ambitious projects in many fields, and show improvements in areas like coding, reasoning, and vision.

My GOAT would never bonk a Chinese spy.

The whole City of Toronto, isometric, and now you can turn it. All 630 km², drawn from four angles. Hit rotate and walk around the CN Tower, or your own street. 🧵 iso.sloptoronto.ca

we have a winner for the most absurd AI startup ever this Peter Thiel backed company uses AI robots to stab fish in the brain so they taste better and last up to 3x longer it's called Shinkei and it modernizes a centuries-old Japanese technique dozens of Michelin-star restaurants are already lining up to buy the fish here's how it works: > the problem is when a fish gets caught, it usually just suffocates on the boat > that stress floods the meat with hormones and lactic acid, which hurts the flavor and texture > the old Japanese fix for this was spiking the fish through the brain right after the catch, which stops all that > but doing it by hand is impractical for most American fishing crews > so Shinkei built a fridge-sized robot that sits right on the boat's deck > the crew feeds in a fish and AI identifies the exact species, so it knows exactly where the brain and gills are > the robot then drives the spike through the brain and cuts near the gills in about 7 seconds > the fish goes into an ice slurry so the blood drains out > so the fish doesn't sit there stressing out, and that tasty flavor and texture is preserved example number 5,392,417 of how AI can be applied to literally anything

Spotify just announced a feature that could either save the music industry or destroy it. Starting as early as Q1 2027, Spotify Premium subscribers will be able to use AI to transform any participating artist’s song into a completely different version. Turn a folk ballad into a dance track. Make a hip-hop song sound acoustic. Generate a cover where one artist’s voice performs another artist’s song. Remix a pop hit into lo-fi study music. Create the version of your favorite song that only existed in your head until now. The remixes will be uploaded directly to Spotify. They’ll appear in playlists. They’ll count toward the original song’s chart position. And most importantly the original artist will earn royalties every time someone listens to a fan-created remix of their work. Spotify has already signed a landmark licensing deal with Universal Music Group the world’s largest record label. Negotiations with Sony Music and Warner Music are ongoing. The feature will cost an additional $4–$5/month on top of Spotify Premium. And analysts project 7–8% of American subscribers will adopt it by 2028. This is not Suno or Udio generating music from text prompts. This is fans remixing real songs by real artists with the artists’ permission, on the world’s largest streaming platform, with royalties flowing back to the creators. A music industry analyst who’s covered streaming economics for 15 years told her team something after reading the licensing agreement that reframed how she understood the future of recorded music. “Spotify just turned 640 million listeners into 640 million potential remixers. The question isn’t whether people will use it they will. The question is whether paying artists for fan remixes solves the economics of streaming or creates a world where the original recording is worth less than the infinite remixes generated from it.” Here’s what Spotify’s AI remix feature actually is, who wins, who loses, and why this might be the most consequential decision in the music industry since the launch of streaming itself 🧵

Apple didn't store a plaintext list of forbidden words in iOS. Instead, researchers discovered a compiled minimal perfect hash table of 32-bit fingerprints buried inside the keyboard's language-model files. When you swipe-type or use autocomplete on an iPhone, the keyboard converts your input into token IDs and evaluates sequences of two to six words. It hashes those candidate n-grams using Bob Jenkins's public-domain hashword() function. If that 32-bit fingerprint matches an entry in the table, the predictive engine refuses to suggest it, or the swipe model actively diverts the word to an alternate match. On November 6, researcher Derik Kauffman published Swipeblocked after exhaustively enumerating candidate word combinations against the table to reverse-engineer the hidden phrases. His conservative estimate is roughly 100,000 entries across 28 languages. In English, the prediction list holds 4,525 phrases and the swipe-typing list holds 2,794. Most entries cover violence, slurs, and sexual content, but the table also silences geopolitical phrases like "Tibetan independence", "United Ireland", "dictator Trump", and Hebrew for "the occupied territories". In testing, swiping "United Ireland" consistently resolved to "United Iceland". You can still type any phrase letter by letter. The suppression only touches autocomplete predictions and gesture paths. Apple told Reuters the system avoids offensive suggestions, but never addressed the political entries or the existence of the table. Apple never published the list because it was never meant to be read. It took brute-forcing the hash table to show what the keyboard was quietly dropping.

hacer el video animado cuesta menos de 10 usd, pero mis amigos de @makingendless se pusieron la 10 y me dieron acceso a su plataforma para que lo haga gratis 😎 hasta ahora la mejor UX que tuve para hacer videos con seedance

This is f*cking insane. Former OpenAI CTO, Mira Murati: "The models we work with today are turn-based. You talk, they talk, then they go off and think, and while they're thinking they're deaf and blind The next ones never stop listening. They take in audio, text and video all the time and answer in 20-millisecond chunks" In just 2 minutes, she shows why the chatbot you type into is about to look like a fax machine Prompt Box → Chatbot → Agents → AI That Works Next To You In Real Time A turn-based model waits for you to finish typing, then goes quiet while it thinks An interaction model sees and hears what you're doing, and can cut in the moment you start going the wrong way Anyone can type a prompt, almost nobody is ready to work with an AI that never stops paying attention Bookmark and watch it today Then read how to set up the agents that will run on models like this in the harness engineering guide below ↓

Introducing Personal Finance in Claude, in beta in the US. Connect the accounts you choose, and Claude answers your money questions in everyday language, from your own balances and transactions. Access is read-only, and Claude can't move, manage, or invest your money.

Pope Leo XIV met with American singer-songwriter Bob Dylan in a private audience on Wednesday, November 4.

We decoupled memory from the model. Living Brain is one persistent brain for ChatGPT, Claude, Cursor, Codex and your own agents. Write in one, retrieve in any. 92.2% on LoCoMo. 464ms median retrieval. Up to 22.6x fewer tokens per question. LLMs have parameters. Living Brain has plasticity.

$1.40 worth of @explabsai on GLM 5.3 uncensored + a day of running a scanner = lots of exposed keys from different services worth $$. Of course I'm not going to use them, but I wanted to see if it was really this easy. It is. All organizations should be using uncensored models to test their own security, because Anthropic and OpenAI won't allow you to do this (without jumping through hoops, at least).

if you're a founder looking to pivot or start something new, don't build another sandbox agent provider. start a construction firm and make it ai native. you'll build a better business and create way more value for the world everyone is crowding into the same few ideas, while the rest of the economy is basically untouched most businesses in the world are old. they were built before software, and most never really adopted it. ai changes that, because it makes both great operations and technical execution much easier than before so the opportunity is no longer selling tools to these industries. it's rebuilding them from scratch, ai native from day one new construction firms. new schools. new hospitals. new clinics, logistics companies, property managers. every old industry is waiting for someone to rebuild it in ai infra you compete with 50 funded teams shipping the same thing every week. in a boring industry you compete with companies that haven't changed in 25 years pick one, obsess over it, and learn the business better than the people in it. the technical part is the easy part now the time has come to improve everything

Since when can I gamble on airbnb?😭 You bet $14 that it’ll rain and you can win $114 a day, absolutely insane odds

Cómo lo hace pasar de largo, por favor. 😮‍💨