Can LLMs be world models in Enterprise? We actually tested it: and the answer is no. While everyone was playing MAPs, we were secretly running something else inside it. WALL-E, a LLM based world model, failed instantly. So today we’re launching CASSANDRA: the first causal world

How an LLM-based world model performs compared to Cassandra💡 This is all the evidence you need: you cannot let an LLM-based model run your enterprise. 📌 Read our blog for full details: https://t.co/B6ikQD9miW

Yann LeCun explains that large language models are trained on about 30 trillion words, representing nearly all public internet text. He says it would take a human over 500,000 years to read that much. But a 4-year-old child sees just as much visual data in their first few years

📌 Check our blog for more details:

If there's one thing that Cassandra showed us: Enterprises need more than LLMs to make strategic decisions. They need: - Causal understanding - Temporal reasoning - Uncertainty modeling - Operational foresight - Resource allocation over time LLM-based world

“human intelligence will not be the upper limit of what’s possible. we’re gonna see AI go far beyond human intelligence.” chief AGI scientist of Google DeepMind believes super intelligence will inevitably surpass humans as hardware and software progress.

Yann LeCun argues there is no such thing as “general intelligence” human or artificial. What we call general intelligence is just our ability to solve problems we’re wired for or can imagine. Humans are great at navigating social life and the real world but terrible at tasks

Google DeepMind CEO Demis Hassabis believes there is virtually no limit to what AI can eventually achieve. "humans as biological information processing systems". He is working on the premise that the entire universe is "computable," meaning that anything that exists can

Some more thoughts about Yann interview: Even if LLMs work great, that's missing the point. Everyone's doing the same thing now. More scale, more data, longer CoT, tweak RL. But the path to get there was completely stochastic. Attention, transformers, scaling laws, RLHF, none of

We recently launched Cassandra, the first causal world model built for business decision making. It beats an LLM-based world model.⚠️ CASSANDRA learns through two tightly-integrated systems: - Deterministic dynamics: LLM-generated code optimized by evolutionary algorithm using

We let an LLM-based World Model (WALL-E) run an amusement park (MAPs). 🎢 Results? - 25% survival rate - 75% bankruptcy rate - Profit collapsed to zero. Turns out LLMs hallucinate business decisions just like they hallucinate facts.

Last month we pitted humans vs AI in enterprise tasks. The results? 👀 Humans outperformed AI. Why? Because LLMs guess probabilities, they don’t model the world ❌ Enter Cassandra - the first causal World Model which enable: - Deterministic logic - Stochastic

Why LLMs cannot be used for business purposes other than as internal tools: When you are an end customer and a service provider tells you that you can use their product, but this product might lie to you at any time or provide a non-working solution, and this is what you get for

Why did we build MAPs? To challenge AI (and humans!) to run a bustling amusement park: design attractions, hire staff, research innovations, and create the most valuable park! 🎢 See how MAPs stack up vs. top benchmarks like ARC-AGI2. Play the game and test it yourself:

Do LLMs really understand? To cut through the noise around AI, we brought in two experts shaping the field: Yann LeCun (@ylecun) and Adam Brown of @GoogleDeepMind. Watch them in conversation with @JannaLevin about the true intelligence of artificial minds. https://t.co/LnzQGPUIwA

Simply tokenizing data won’t drive impact. Building comprehensive world models will.

In the age of AI, 2026 will be the year of physical intelligence & 3D AI world models, complementing today’s GenAI systems, which are now at the proto-AGI level. This will then jumpstart embodied intelligence in the form of humanoid robots, & the world will never be the same.

A year ago, we verified a preview of an unreleased version of @OpenAI o3 (High) that scored 88% on ARC-AGI-1 at est. $4.5k/task Today, we’ve verified a new GPT-5.2 Pro (X-High) SOTA score of 90.5% at $11.64/task This represents a ~390X efficiency improvement in one year

LLMs vs World Models: How different AI models score across key qualities of an ‘AI CEO’ See the full breakdown in our blog: ➡️ https://t.co/B6ikQD9U8u #EnterpriseAI #LLMs

Yann LeCun (Chief AI Scientist, Meta, @ylecun), @PimDeWitte (CEO, General Intuition), and Aude Durand (Kyutai, @aude_drn), talk about world models, embodied agents, Yann's new company, and the limitations of LLMs 0:00 - Introduction to World Models 5:00 - Why World Models,