Salesforce AI Research introduces MCP-Universe: the first benchmark to truly test LLM agents in real-world scenarios with live Model Context Protocol servers. https://t.co/G85e4qGgW4

OpenAI's GPT-5 tested for spatial intelligence. New study evaluates GPT-5 & other MLLMs on 8 benchmarks. Shows unprecedented strength, but still falls short of human performance. https://t.co/FbYlfcDPQt

Tinker delivers high-fidelity 3D editing from sparse inputs. It achieves multi-view consistent edits from just one or two images, without any per-scene finetuning. A truly scalable, zero-shot approach to 3D content creation. https://t.co/uEk9tcMVQB

Grok 2 is out on Hugging Face https://t.co/2nwv6DlzlM

Qwen-Code weekly release (v0.0.8) : ✨ Deep VS Code Integration: Get context-aware suggestions & inline diffs directly in your editor! Initialize with /ide and supercharge your workflow. 🔌 Enhanced MCP Support: Add, remove, list MCP servers via CLI (qwen mcp add|remove|list), plus support for MCP roots & multimodal tool outputs (images, audio, and more!). 📱 Responsive Terminal UI: Now adapts gracefully to narrow terminals. 🔍 Reverse Search: Press Ctrl+R in shell mode to instantly search command history. 🗜️ Customizable Context Compression: Set your threshold in settings.json. 📂 Multi-directory Support: Auto-load multiple dirs at startup. 🔢 Hide Line Numbers: Set "showLineNumbers": false for cleaner copy-paste. 🔕 Plus: Disable update nudges with "disableUpdateNag": true 📌More details: https://t.co/qqwj5nAger Upgrade now and level up your coding experience 💻🔥

It is interesting that the new @deepseek_ai v3.1 is trained using the UE8M0 FP8 scale data format which is logarithmic number system. Our multiplicative weights update (Madam) for training in that format was done several years ago while at @nvidia It yields maximum hardware efficiency with no accuracy loss https://t.co/LaRlF9Nesk

There is unreasonable alpha in building your own annotation app - Love this example by @ChrisLovejoy_ . Vendors often fight us when we say it, but @sh_reya and I stand our ground for this exact reason. > Compare the images of a generic trace viewer vs the one he built. It is so much easier for doctors to get the context of what is going on in the latter, and they can provide helpful labeling without nearly as much typing. See Nick's post below for full article.

For folks looking to admin a few boxen without amping up the complexity, consider using a cluster ssh tool like ClusterShell or Fabric. It just sends ssh commands to groups of boxen in parallel and shows you the consolidated result. E.g https://t.co/pQAYI9o4B8 https://t.co/SOMoGXHc0L

How much better is modern bluetooth audio compared to the past? The answer is a lot…and also not at all. It all depends on the mode. One-way, with modern codecs (music, etc) is quite good. Headset mode (bidirectional, think zoom calls) is still awful: https://t.co/SDzOiZG26y

We are thrilled to introduce the Seed-OSS family of open-source LLMs, developed by ByteDance's Seed Team. GitHub: https://t.co/lUNRuigqMA HuggingFace: https://t.co/1WuQHpGcIo Feel free to try it out and share your feedback! https://t.co/Q8kYgO1uT2

+4 more

Intern-S1-mini 🔥 lightweight open multimodal reasoning model by @intern_lm https://t.co/aTXFjyv6Q5 ✨ Efficient 8B LLM + 0.3B vision encoder ✨ Apache 2.0 ✨ 5T multimodal pretraining, 50%+ in scientific domains ✨ Dynamic tokenizer for molecules & protein sequences

Intern-S1-mini 🔥 lightweight open multimodal reasoning model by @intern_lm https://t.co/aTXFjyv6Q5 ✨ Efficient 8B LLM + 0.3B vision encoder ✨ Apache 2.0 ✨ 5T multimodal pretraining, 50%+ in scientific domains ✨ Dynamic tokenizer for molecules & protein sequences

Introducing Command A Reasoning, our most advanced model for enterprise reasoning tasks. https://t.co/p0UjXCISle

Introducing Command A Reasoning, our most advanced model for enterprise reasoning tasks. https://t.co/p0UjXCISle

Major TOM @GoogleDeepMind's AlphaEarth Embeddings are now on @huggingface! 🚀 A new 6 TB prototype dataset for the community. Get it here: https://t.co/jMlDChvuI3 #EarthEngine #AlphaEarth

Qwen Image LoRAs are growing fast up on the @huggingface Hub, and now they can be inferenced them via @FAL on the API or directly via the model page https://t.co/fVK8tBAeOj

🏓 After mastering chess, xylophone and Jenga towers, Reachy 2 is now taking on ping-pong! Low-latency teleoperation allows the operator to react quickly enough to return the ball. https://t.co/UZSCNaJ0TA

After a few days of experimentation, I think I’ve cracked Qwen-Image training! Wait until you see what Qwen-Image LoRAs unlocks 🔥 https://t.co/y5clNv8kts

Three controlnets for Qwen Image are now integrated in Comfy Haven’t tested yet https://t.co/tYWdi75UEE

Three controlnets for Qwen Image are now integrated in Comfy Haven’t tested yet https://t.co/tYWdi75UEE

🚀 First week at @huggingface: shipped improvements to The Ultra Scale Playbook! ✅ Dark mode ✅ Some mobile responsiveness ✅ Performance fixes Your complete guide to scaling #LLMs in 2025 👇 https://t.co/MHnkmUR9OB

Imagine having a ping pong robot! 🏓 Researchers and developers building physical AI: meet Reachy 2 from @pollenrobotics, an open-source, humanoid robot for real-world experimentation. It’s a bimanual mobile manipulator: each 7-DOF arm mimics human proportions and can lift up to 3 kg, giving dexterity for object handling. It can be controlled with Python and ROS2 Humble, or go straight into VR teleoperation, use a headset to move Reachy’s arms, hands, and head, and see through its cameras as if you’re in the robot’s own body. Want it to move around? A mobile base with three omnidirectional wheels, rich sensors, and LiDAR lets Reachy 2 navigate and explore its surroundings smoothly. 🗺️ Under the hood, it’s powered by a CPU system that’s ready for machine learning, perfect for loading AI frameworks and testing new models from @huggingface directly on the robot. Keep making robots more, and more accessible Pollen team! ... and keep making more open source models to make robots more mainstream @ClementDelangue!

Thanks to Saba from the XetHub team at @HuggingFace, Trackio now supports logging images. Again, completely for free! What should we add next? https://t.co/rlOhD26NfF

Wrote an FA3 attention processor for @Alibaba_Qwen Image using the 🤗 Kernels library. The process is so enjoyable! Stuff cooking stuff coming 🥠 https://t.co/qVtQkwoB6o https://t.co/nkkvP67xyz

I am speaking at #AIDev on 29th August in Amsterdam, covering open video generation models. Come say hi 🤗 Also, happy to meet with folks in 🇳🇱 working in similar areas. Reach out if interested :) Links and more are in comments. https://t.co/gjsDcyevA3

xAI just released Grok 2 on Hugging Face. This massive 500GB model, a core part of xAI's 2024 work, is now openly available to push the boundaries of AI research. https://t.co/nYEAMLP2zv

W&B Fully Connected 2025. Learned about a lot of cool technologies and ppl. Was super fun! https://t.co/fb4Otw45XK

+2 more

W&B Fully Connected 2025. Learned about a lot of cool technologies and ppl. Was super fun! https://t.co/fb4Otw45XK

+2 more

@manikxpardan @MarvellTech Where to find him: GitHub: https://t.co/KVXIMwh4gm X: https://t.co/HgJnSiuz15 LinkedIn: https://t.co/MwctuGALri

⌛ VS Code Live starts in 10 minutes! Today @ReynaldAdolphe chats with @samidip and @kris_mariyanov about how to utilize @Telerik & KendoUI AI Coding Assistants to streamline your development experience. Join us to chat live with the team ▶️ https://t.co/UFK1rz1O0s https://t.co/qEgt56bsfF

@ns123abc https://t.co/RRpw3ioDoX

Very proud to be a Partner of @sparklabsgroup, recognized by @TIME and @StatistaCharts in the top 10% of venture capital firms in the U.S. 🚀 👉 https://t.co/VKJ26hTumb @bernardmoon