Major technological advances and scientific discoveries pushing the boundaries of what's possible.
Most AI video generators create a single, disconnected clip. PAN, a new model from MBZUAI, is different. It acts like an interactive simulator, maintaining a persistent world that reacts to your commands over time.
We all want faster AI, but not at the cost of quality. NVIDIA's new TiDAR architecture is a clever hybrid approach that drafts text like a diffusion model and verifies it like an autoregressive one, all at once, leading to a massive speed boost.
Just hours after OpenAI's latest update, Baidu dropped its own bombshell: ERNIE 5.0. We break down how this new AI model claims to beat GPT-5 in key areas, what it costs, and why it's a huge move in the global AI race.
We usually think of AI as a chatbot, but what happens when it gets a body? Anthropic just connected its Claude AI to a robot dog, and the results are a fascinating glimpse into our physical future.
A new 1.5 billion parameter AI model from Weibo is making waves. Trained for just $7,800, VibeThinker-1.5B is outperforming models hundreds of times its size on complex reasoning tasks. Here's how they did it and why it matters.
Google is about to drop a new AI image model called Nano Banana 2, and it's not just another update. Early peeks suggest it's hyper-realistic, understands text perfectly, and even corrects its own mistakes before you see them.
AI voices are getting a major upgrade. Maya1 is a new open-source model that generates incredibly expressive, controllable speech from text, and the best part? It runs on a single gaming GPU.
Forget what you know about speech recognition. Meta just open-sourced Omnilingual ASR, an AI model that supports over 1,600 languages, dwarfing OpenAI's Whisper. Here's why this is one of the most important AI releases of the year.
Frustrated with AI agents that mis-click? A new model called Gelato-30B-A3B is here to fix that. We break down how this specialized AI, trained on a super-clean dataset, is making computer-use agents smarter and more reliable.
AI agents are getting smarter, but they're also incredibly expensive to run. Anthropic has a brilliant new fix that stops AI from "thinking out loud" and starts letting it write code, slashing token usage by over 98%.
AI models often forget old information when they learn new things, a problem called catastrophic forgetting. Google's new "Nested Learning" approach rethinks models as a collection of smaller systems that learn at different speeds, much like a human brain.
An Indian startup just launched an AI video generator that plugs directly into apps. This isn't just another cool tech demo—it's a sign that the power to create movies is shifting from Hollywood studios to anyone with a few lines of code.