Model Context Protocol server for web search without API keys

Qwen-Image-Edit-2509-Light_restoration app

Qwen Image Edit w/ Camera Control is wild 🤯 Quickly rotate the camera, switch between bird's eye and worm's eye views using just clicks. Here's how plus 7 wild examples:👇

Here are some slides I made for a short tutorial on Non-Euclidean gradient descent and these variants: docs.google.com/presentation…

When you also throw truncation in the mix, that gives a lot of possible variants of Muon. We tried out many, here are some of the winners (including Scion)

this is great

So much of biomedical research rests on tacit knowledge—unwritten and invisible to LLMs—causing agents to underperform. The Biomni Open Know-How Catalogue bridges that gap: a curated library of human expertise that Biomni agents use intelligently on the fly, with dramatic gains. Build together: biomni.stanford.edu/blog/bio…

Someone from xAI reached out and asked me to retest grok-4-fast, because they've improved the injected system prompts. Huge improvement! grok-4-fast-reasoning: 77.5% -> 94.1% grok-4-fast-non-reasoning: 77.9 -> 97.9% I really appreciate that xAI takes this topic seriously.

💥The future of protein AI isn't 3D, it's 4D💥 PTraj-Diff = SE(3)-diffusion (geometry) + BERT encoder (time) A custom TPA (90% mem reduction) makes it feasible. The result: high-quality, long-range trajectories seeded directly from static AF3 folds

If you are working on real-to-sim, simulating digital twins, and policy evaluation, you should check out our fully open-sourced code base. Lots of handy tools for building Gaussian Splatting simulators and interacting with it! github.com/kywind/real2sim-e… Will continue to be maintained and expanded.

verifiers v0.1.7 is released 🚀 this one's all about making RL training and experimentation waaaay easier: - single-command installation for prime-rl - single-command training w/ unified configs - overhauled vf.RLTrainer for hacking on new algorithms quick demo + links below :)

Finished "Neuroscience" lecture series 🥳 started on "Manifold learning", same teacher aka "dimensionality reduction" I always thought in ML when we say "this data is 1000s of dimensions" etc, that this is the accurate dims of representation of it and dim reduction is an approximation so our eyes can perceive it at some level But actually, the true representation of a 1000 dimensional data can be 50 dimensions perhapsl or even 3 or 2 so those 1000s Dims are just the space that represents it but that doesn't mean that is the most meaningful representation this is a quick introductory course and I highly recommend it. only 7 short lectures She really teaches well and the lecture talks about methods as well like t-SNE, UMAP, PCA etc I will put a link in comment

🚨 Big news in the AI world! Baidu’s ERNIE-5.0-Preview-1022 just scored 1432 on the latest LMArena Text Leaderboard, making it #1 in China and #2 globally! The model really stands out in creative writing, complex reasoning, and following instructions. The ERNIE-5.0 foundation model is reportedly set to launch at Baidu World 2025 on November 13. The ERNIE series has gone through years f continuous improvement — from multimodal models like ERNIE 4.5 and 4.5 Turbo to deep-thinking models ERNIE X1, X1 Turbo, and X1.1 — consistently leading the way in Chinese large language models. 👉 Check out the leaderboard here: lmarena.ai/leaderboard/text #AI #LLM #ERNIE5 #BaiduWorld2025 #MachineLearning #Innovation

Hope.

Big Data Pipeline Cheatsheet for AWS, Azure, and Google Cloud

Squidiff: a diffusion-based model to predict transcriptome response to perturbations. nature.com/articles/s41592-0…

Today’s edition of my newsletter just went out. 🔗 rohan-paul.com/p/alibaba-bac… Consider subscribing, its free, and I write it everyday. 🇨🇳 Alibaba-backed Moonshot releases Kimi K2 Thinking 🏆 New Stanford+Oxford and other top university study just dropped highlighting flaws in AI benchmarking. ⚖️ Amazon has sued Perplexity alleging that Perplexity’s Comet browser runs an AI shopping agent that logs into Amazon with a user’s credentials. 📈 Google is finally rolling out its most powerful Ironwood AI chip, first introduced in April, taking aim at Nvidia in the coming weeks. ⚖️ 📈 Microsoft Research released Magentic Marketplace, an open-source environment that lets people test how LLM agents buy, sell, negotiate, and pay at scale, revealing real issues in discovery, fairness, and safety. 📈 Edison Scientific launched Kosmos, an autonomous AI researcher that reads literature, writes and runs code, tests ideas.

Named Entity Recognition is a tool that helps you pick out important terms in text. It's helpful for extracting meaningful insights from large bodies of text, for example. In this tutorial, Manish teaches you how it works by building a news analyzer that uses a transformer-based NER model to grab data from a live RSS feed. freecodecamp.org/news/extrac…

It's rare nowadays to find something that is intuitively important and not yet done well by any major language models. But *precisely aggregating lots of information over long contexts* is one of those things. Our new benchmark Oolong tests this ability, see the 🧵 for more!

China will surpass the US in AI. They just dropped an open-source model that beats GPT-5 and Claude on reasoning benchmarks. Kimi K2 Thinking from Moonshot AI posted 44.9% on Humanity's Last Exam when GPT-5 only hit 33%. Beat Claude Sonnet 4.5 on competitive programming. Crushed both on agentic search and coding tasks. This isn't some research lab demo. The model executes 200-300 sequential tool calls without human interference, has a 256K context window, and went live on their platform yesterday with full API access. Moonshot AI is a 2-year-old startup founded by ex-Tsinghua researchers. No state backing. No CCP money printing their runway. Just 200 engineers in Beijing building test-time scaling that actually works. The model weights are on Hugging Face right now. Open source. Anyone can run it. While American AI labs fight over safety theater and regulatory capture, China is shipping production-grade reasoning models faster than we can benchmark them. DeepSeek taught them how to train cheap, now Moonshot showed them how to scale reasoning without burning $100M per training run. The gap is closing faster than anyone wants to admit.