The first fantastic paper on scaling RL with LLMs just dropped. I strongly recommend taking a look and will be sharing more thoughts on the blog soon. The Art of Scaling Reinforcement Learning Compute for LLMs Khatri & Madaan et al.

New post! This time, about the current state of Long Context Evaluation. I discuss existing benchmarks, what makes a good long context eval, what's missing from existing ones and introduce a new one - LongCodeEdit :)

New: Internal Amazon documents warn AI startups are delaying and diversifying their cloud spending. There are also internal concerns about AWS’ pricing and lagging reputation in AI. businessinsider.com/amazon-ai-star…

We found a new way to get language models to reason. 🤯 No RL, no training, no verifiers, no prompting. ❌ With better sampling, base models can achieve single-shot reasoning on par with (or better than!) GRPO while avoiding its characteristic loss in generation diversity.

We evaluated Claude Haiku 4.5 on several benchmarks. Even with reasoning disabled, Haiku 4.5 performs similarly or better than early lightweight reasoning models, like o1-mini.

Alpha Arena is LIVE 6 AI models trading $10K each, fully autonomously Real money. Real markets. Real benchmark. Who's your money on? Link below

Hi! Does anyone know of a simple but usable (i.e., not O(N^2)) implementation of PointNet++ in Pytorch? This is what I mean by "simple" github.com/thuml/Neural-S… but alas, this is PointNet, not PointNet++

My pleasure to come on Dwarkesh last week, I thought the questions and conversation were really good. I re-watched the pod just now too. First of all, yes I know, and I'm sorry that I speak so fast :). It's to my detriment because sometimes my speaking thread out-executes my

Shocking: MoM growth slows down as you penetrate the market More context: 15% of 800M is 120M. Working population: 46M (GER) + 24M (ESP) + 38M (ITA) + 30M (FRA) = 138M (total: 255M); so they captured the maj of the available market already

new whale dropped

So longer read of DeepSeek-OCR It’s an engineering achievement. It has been suspected for a while that VLM/OCR models could be significantly smaller. The pre-VLM state of the art, Google Cloud OCR would not be more than a 100m model. More recently, relatively small open weights

Tomorrow is a special day for the AI Studio team. Since May, we have been heads down building a brand new AI vibe coding experience to accelerate the path from prompt to production with Gemini. Can’t wait to show you all :)

Introducing Qwen3-VL-2B and Qwen3-VL-32B! From edge to cloud, these dense powerhouses deliver ultimate performance per GPU memory, packing the full capabilities of Qwen3-VL into compact and scalable forms. 🔥 Qwen3-VL-32B outperforms GPT-5 mini & Claude 4 Sonnet across STEM,

Below is a deep dive into why self play works for two-player zero-sum (2p0s) games like Go/Poker/Starcraft but is so much harder to use in "real world" domains. tl;dr: self play converges to minimax in 2p0s games, and minimax is really useful in those games. Every finite 2p0s

As part of our recent work on memory layer architectures, I wrote up some of my thoughts on the continual learning problem broadly: Blog post: jessylin.com/2025/10/20/con… Some of the exposition goes beyond mem layers, so I thought it'd be useful to highlight separately:

I used ChatGPT to solve an open problem in convex optimization. *Part I* (1/N)

Today, we are open-sourcing Hunyuan World 1.1 (WorldMirror), a universal feed-forward 3D reconstruction model. 🚀🚀🚀   While our previously released Hunyuan World 1.0 (open-sourced, lite version deployable on consumer GPUs) focused on generating 3D worlds from text or