Omar Saadoun

@joaomdmoura seems to be a general practice now

damn someone is nerfing opus pretty hard it feels like

CEO obsessed with token maxxing

Yann LeCun was right the entire time. And generative AI might be a dead end. For the last three years, the entire industry has been obsessed with building bigger LLMs. Trillions of parameters. Billions in compute. The theory was simple: if you make the model big enough, it will eventually understand how the world works. Yann LeCun said that was stupid. He argued that generative AI is fundamentally inefficient. When an AI predicts the next word, or generates the next pixel, it wastes massive amounts of compute on surface-level details. It memorizes patterns instead of learning the actual physics of reality. He proposed a different path: JEPA (Joint-Embedding Predictive Architecture). Instead of forcing the AI to paint the world pixel by pixel, JEPA forces it to predict abstract concepts. It predicts what happens next in a compressed "thought space." But for years, JEPA had a fatal flaw. It suffered from "representation collapse." Because the AI was allowed to simplify reality, it would cheat. It would simplify everything so much that a dog, a car, and a human all looked identical. It learned nothing. To fix it, engineers had to use insanely complex hacks, frozen encoders, and massive compute overheads. Until today. Researchers just dropped a paper called "LeWorldModel" (LeWM). They completely solved the collapse problem. They replaced the complex engineering hacks with a single, elegant mathematical regularizer. It forces the AI's internal "thoughts" into a perfect Gaussian distribution. The AI can no longer cheat. It is forced to understand the physical structure of reality to make its predictions. The results completely rewrite the economics of AI. LeWM didn't need a massive, centralized supercomputer. It has just 15 million parameters. It trains on a single, standard GPU in a few hours. Yet it plans 48x faster than massive foundation world models. It intrinsically understands physics. It instantly detects impossible events. We spent billions trying to force massive server farms to memorize the internet. Now, a tiny model running locally on a single graphics card is actually learning how the real world works.

@claudeai You shouldn't be biased like that, we know :)

@HeyGen Rough first test of Tina, the Customer Retention Engine. customerretentionengine.com

@osaadoun @osaadoun DM sent! Let us know what you build 🛠️

We built our launch video in Claude Code using HyperFrames. Now it's yours. Open source, agent-native framework. HTML to MP4. $ npx skills add heygen-com/hyperframes RT + Comment "HyperFrames" to get the full source code of this launch video (must follow)

@HeyGen Did you send it? I don't see it.

@mauricioclausen @AntelDeTodos Prendele a starlink, ni las tormentas lo afectan!

@AntelDeTodos Semana y media y sigo sin internet @AntelDeTodos! Así no hay como no pensar en Starlink.

@AntelDeTodos estoy sin internet desde la madrugada. No funcionaba reclamo la app, pude luego en la web. Pero no tengo ninguna actualización del estado y previsión de reparación! Llamo por tel y la máquina dice vea en la web u corta! Pésimo servicio, preciso saber un plazo!

@claudeai @grok compare Claude managed Agents with @crewAIInc from a technical standpoint, production readiness and costs at scale

Introducing Claude Managed Agents: everything you need to build and deploy agents at scale. It pairs an agent harness tuned for performance with production infrastructure, so you can go from prototype to launch in days. Now in public beta on the Claude Platform.

Only today, I created a couple of Cursor Cloud Agents that review my product documentation, check the code base and find product gaps, propose new features code them on both front and backend for me to review later, another agent to look for errors on the deployed versions and fix them. I also created a PaperclipAI company that runs all the marketing team, plans, deliverables and is executing non-stop (or as long as they have enough credits lol). I feel like I'm losing time by writing this, I should continue building (and spending AI Credits!!).

That's why we need more agents? 😇

Well, who saw this coming? "AI is particularly useful in beef processing because of the variability involved, Schattenmann said. “It’s not like assembling a car, right? Every Toyota Camry looks the same and it’s all automated. Every cow is slightly different.”"

🚨 Do you understand what happened in the last 12 hours? > Goldman Sachs said $450 billion in AI investment contributed "basically zero" to US economic growth.. the Chief Economist said it on record.. > Uber dropped $1.25 billion on Rivian to build 50,000 robotaxis.. the drivers who built the company are about to be replaced by the cars.. > Jeff Bezos started raising $100 billion to buy manufacturing companies and automate them with AI.. he's not building factories.. he's buying them to empty them.. > DoorDash launched an app that pays people to film themselves doing chores.. not for content.. for AI training data.. so robots can learn to replace them.. > OpenAI bought Astral.. the open source Python tools that millions of developers loved.. bolted it onto their coding agent.. the community is furious.. > Cursor dropped Composer 2.. their own AI model that beats Claude on coding benchmarks.. a 50-person startup outperformed a $30 billion lab.. > Hollywood resurrected Val Kilmer with AI to star in a new movie.. he died of cancer.. never filmed a single scene.. they built a digital version of him.. > HSBC is planning deep job cuts using AI to gut their middle and back offices.. > UK unemployment hit 5.2%.. AI hasn't even started firing people yet.. > Fortune 500 updated the AI price tag to $4.5 trillion.. 93% of US jobs vulnerable to disruption.. > Nvidia GTC is happening right now.. Jensen closed his keynote with singing robots and a digital clone of himself.. the man selling the shovels is also selling the future.. all of this happened in ONE day.. a single Thursday.. if you're not following me you're finding out about this 48 hours late from someone who read my post..

@Scobleizer It would be great to have some way of trying without compromising 2K/month. I've seen the CL SDK but still could be awesome to have a cheaper sub or paid trial (maybe for a shorter period of time) .

Human neurons running in the cloud. The matrix is here. (I was the first to reveal this company back in 2023, still one of the most exciting, scary, and wild things I've ever seen).

🤯🤯

When your coding agents are running for 20 min non stop for a feature you know you are into something big.

So agree with this thinking from @aleximm : "The companies that win in this environment will be the ones delivering genuine value, not the ones that built the highest walls around their customer base." Real value is what matters.

Tell me how many AI credits your product is spending and I'll tell you how AI Native it is.

We are living the Agentic thinking era. It’s not anymore how do I do this task but how do I get an agent to do it for me. Just don’t overthink it or you might end up hammering everything because you now have a hammer.

@_fiatisabubble You better check who's holding tokens before making such statement, that's a team wallet used for security purposes to send tokens to the vesting contract. I usually don't like to tell potential attacker where to look. Do you?

crazy_22.near dumping $ITLX on vested presale holders 🙈