🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

More great work from Toby Simonds Tufalabs

1/ Gemini 2.5 is here, and it’s our most intelligent AI model ever. Our first 2.5 model, Gemini 2.5 Pro Experimental is a state-of-the-art thinking model, leading in a wide range of benchmarks – with impressive improvements in enhanced reasoning and coding and now #1 on

WE HAVE A NEW BEST MODEL IN THE WORLD! GEMINI 2.5 IS #1 ON LIVEBENCH

Last month we launched our Anthropic Economic Index, to help track the effect of AI on labor markets and the economy. Today, we’re releasing the second research report from the Index, and sharing several more datasets based on anonymized Claude usage data.

So tempting...

We are hiring full-time senior video editors and motion graphics artists in the UK to work in our offices near London. We are making some of the most interesting technical content on YouTube, it will be exciting! Hit me up or refer friends 🙏

AI beliefs & reward functions ≠ ours. While simple games (Go, Chess) have clear win/lose rewards, complex real-world situations don't. "Reward is all you need" is partly true, but where does the reward function come from? This is Dr. Jeff Beck from Noumenal Labs

Why does linear regression minimize squared error? I always thought it was just computationally convenient compared to absolute values. But then I discovered it naturally emerges from maximum likelihood estimation with Gaussian noise assumptions 🤯 Same with regularization—L2/L1

Amazing work from Meta. We will probably see a variety of competitive drops in the next few weeks.

.Bryan Landers made this sweet graphic for us It's a public, internal tool we're hosting here: arcprize.org/2025-sota Not guaranteed to be up to date (we're getting there though)

Amazing application of TTT to video generation.

In a recent discussion, Prof. Kevin Ellis and Dr. Zenna Tavares of Basis explored pathways toward artificial intelligence capable of more human-like learning, focusing on acquiring abstract knowledge from limited experience through active interaction. 🧵👇

Ouch... Maybe the reasoning version in the future will fare better.

.Isaac Liao has open sourced his "ARC-AGI Without Pretraining" notebook on Kaggle You can use it today and enter ARC Prize 2025 It currently scores 4.17% on ARC-AGI-2 (5th place) Amazing mid-year sharing and contribution Thank you Isaac

Creating a Professional Voice Clone (PVC) of your own voice allows you to produce high-quality voiceovers that sound exactly like you. Today, our team shipped a brand new version of the PVC creation flow that makes it much easier to create a perfect clone of your voice.

Meet Codex CLI—an open-source local coding agent that turns natural language into working code. Tell Codex CLI what to build, fix, or explain, then watch it bring your ideas to life.

Our new OpenAI o3 and o4-mini models further confirm that scaling inference improves intelligence, and that scaling RL shifts up the whole compute vs. intelligence curve. There is still a lot of room to scale both of these further.

Re-testing released o3 on ARC-AGI-1 will take a day or two. Because today's release is a materially different system, we are re-labeling our past reported results as "preview": o3-preview (low): 75.7%, $200/task o3-preview (high): 87.5%, $34.4k/task Above uses o1 pro pricing

Google's new Quantization-Aware Training (QAT) models shrink Gemma 3's VRAM needs by up to 75% — without killing performance. That means: ✅ Run Gemma 3 27B on a 3090 GPU ✅ Use 12B on a laptop ✅ Deploy 1B on... basically a potato Open-source, Hugging Face ready,

Let’s start with Grok 3 Mini. When we set out to build a fast, affordable mini model, we knew it would be good but even we didn’t expect it to be this good. Some highlights: - Grok 3 Mini tops the leaderboards on graduate-level STEM, math, and coding, outcompeting flagship