It's time to solve one of the biggest challenges in vision-language models. Today's multimodal models can look at a biomedical image and give a convincing, sometimes correct, answer without actually understanding what they're seeing. A model might answer the medical question correctly while failing to identify basic visual context like the imaging modality, body part, specimen, or stain. So, I'm teaming up with Stanford to solve this problem. You should too. The MMBU Challenge tests whether models can actually recognize, localize, and understand what is in biomedical images, not just arrive at the right answer. 🏆 3 tracks 💻 $100K+ in compute and prizes 📅 Oct 1 to Dec 31 ⏰ Registration closes today

Snap just entered the agentic race. They just introduced SPECS Intelligence, a copilot that spans Mac, iPhone, and AR glasses. The hard part isn’t connecting email, calendar, and notes. It’s turning that firehose into a continuously updated model of your life. The system has to: • resolve identities across apps • map events to the right project, trip, or relationship • track what changed and what’s stale • isolate work and personal context • rank what matters now • decide when to interrupt • sync state across devices • gate actions behind approval Snap’s answer is a structured context layer built around three primitives: > Portrait: who you are, who matters, and your patterns. > Corners: scoped contexts for work, family, travel, health, and relationships. > Goals: persistent objectives that give individual events longer-term meaning. Instead of treating memory as one giant retrieval pool, @specs can organize context by domain, update it as new events arrive, and surface the relevant slice when needed. And Snap has one unusual advantage, they own a device in your field of view. If agent quality depends on context, glasses could eventually give Snap a source of physical-world context that software-only agents don’t have.

Zuck is back. Meta didn't just add another reasoning model to the market. It shipped a production stack for AI agents. Muse Spark 1.1 combines in a single system: - 1M-token context window - parallel sub-agents - computer use - tool orchestration - context compaction - OpenAI-compatible API The infrastructure needed to build long-running, production-grade agents is becoming part of the model itself instead of something developers have to stitch together. Available via the new Meta Model API, now in public preview.

Just read LeCun's latest paper. His team trained the first world model that can't collapse. Let me explain why this matters. It's called LeWorldModel. World models predict what happens next physically. Objects moving, falling, colliding. That's the base layer for robots that plan, cars that simulate before they steer, any AI that acts in reality instead of just talking about it. The catch is nobody could train these reliably. The models kept cheating. They'd map every input to the same output. Like a weather app stuck on "sunny" forever. Technically predicting. Completely useless. So teams piled on fixes. Frozen encoders, stop-gradient hacks, 6+ loss hyperparameters. A fragile stack too brittle for production. This team asked a different question. What if you make collapse mathematically impossible? An encoder turns each video frame into a small vector. A predictor takes that vector plus an action and guesses the next one. First loss: how wrong was the guess. Second loss: a regularizer called SIGReg that checks if vectors spread out like a bell curve. If they start looking the same, the loss spikes. The model can't cheat because the math won't let it. That simplicity is what makes the results possible. Six hyperparameters became one. 15M parameters. Trains on one GPU in hours. Plans 48x faster. Encodes with ~200x fewer tokens. Open-source. I could run this on my own hardware. Which changes who gets to build physical AI. Not just big labs anymore. Any team, any startup, any grad student. LeCun has pushed JEPA as the path forward. The criticism was always training instability. This paper removes that objection. Two directions compete in AI right now. Bigger LLMs with more compute. Or small models learning physics from raw pixels.

Every foundation model you've ever used has the same bug. It just got fixed. Since 2015, every deep network has been built the same way: each layer does some computation, adds its result to a running total, and passes it forward. Simple. But there's a problem, by layer 100, the signal from any single layer is buried under the sum of everything else. Each new layer matters less and less. Nobody fixed this because it worked well enough. Moonshot AI just changed that. Their new method, Attention Residuals, lets each layer look back at all previous layers and choose which ones actually matter right now. Instead of a blind running total, you get selective retrieval. The analogy: imagine writing an essay where every draft gets merged into one document automatically. By draft 50, your latest edits are invisible. AttnRes lets you keep every draft separate and pull from whichever ones you need. What this fixes: 1. Deeper layers no longer get drowned out 2. Training becomes more stable across the whole network 3. The model uses its own depth more efficiently To make it practical at scale, they group layers into blocks and attend over block summaries instead of every single layer. Overhead at inference: less than 2%. The result: 25% less compute to reach the same performance. Tested on a 48B-parameter model. Holds across sizes. Residual connections have been invisible plumbing for a decade. Now they're becoming dynamic. The next generation of models won't just pass through their own layers, they'll search them.

Yann just bet a billion dollars that the entire industry is building on the wrong foundation. Large language models predict the next word. They're trained on text, so they understand language. But the real world isn't made of words. It's made of continuous sensor data: camera feeds, touch, sound. And most of that data is unpredictable. You can't predict every pixel in a video the way you predict the next token in a sentence. Generative models fail here because they try to predict everything, including noise. AMI Labs is building world models using JEPA (a method LeCun proposed in 2022 that learns abstract representations of reality and predicts in that compressed space, not in raw pixels). Action-conditioned versions let AI simulate the consequences of actions before taking them. That's not generation. That's understanding. This unlocks AI that can operate in the physical world without hallucinating: 1. Robotics that plans multi-step actions 2. Healthcare devices where errors kill patients 3. Industrial process control under safety constraints 4. Wearables that adapt to real-time sensor input If JEPA works at scale, the next wave of AI companies won't fine-tune LLMs. They'll train world models on sensor data. LeCun's CEO already predicts every startup will rebrand as a "world model company" within six months. The architecture war is starting.