https://t.co/QwzTabilUn

One of the hardest problems in robotics isn’t movement — it’s manipulation. This demo shows adaptive switching between left and right robotic hands, allowing the system to decide which hand should take over a task in real time. Human-like dexterity is what will finally bring AI from software into the physical world. 🤖 via Takayuki Yamazaki #Robotics #AI #Humanoid #Automation

Exploring cozy worlds 🍀 24 million splats Take a walk through it right in your browser ↓ https://t.co/an3XR7vSVT

Exploring cozy worlds 🍀 24 million splats Take a walk through it right in your browser ↓ https://t.co/an3XR7vSVT

Try It Here 🌎 https://t.co/QxpvdlBjss

Try It Here 🌎 https://t.co/QxpvdlBjss

🔥 Meet dots.ocr-1.5: 3B OCR model from Rednote-hilab , SOTA multilingual document parsing, virtually any writing system. 📊 Elo 1089 on olmOCR-Bench, 1157 on XDocParse — above GLM-OCR, and PaddleOCR-VL-1.5 📄 OmniDocBench text edit 0.031, beats Qwen3-VL-235B (0.069) and Gemini 2.5 Pro (0.075) 🎨 SVG code output for charts, diagrams, and chemical formulas 🌐 Web parsing, scene text spotting, and object counting included ⚡ vLLM supported, single GPU 🤖 Model: https://t.co/aTqXUALT5u 🔗 GitHub: https://t.co/nYtek81lm4 🎠 Demo: https://t.co/q5RBPWpdpS

+5 more

Ready to deploy AI agents? NVIDIA NemoClaw simplifies running @openclaw always-on assistants with a single command. 🦞 Deploy claws more safely ✨ Run any coding agent 🌍 Deploy anywhere Try now with a free NVIDIA Brev Launchable 🔗 https://t.co/eAbqav9A4x https://t.co/V1OcqXxvOE

Hallucinations in large language models are often blamed on bad data. But the deeper issue may be the architecture itself. These systems are designed to predict the most plausible next word, not to verify truth. If hallucinations are part of the design, the real challenge is learning where and how we can safely rely on them. https://t.co/c0OLuQ0M8y @TDataScience

🧠 Want to give your coding agent new capabilities? Use Agent Skills in @code! Agent skills let you package instructions and resources into reusable capabilities for your agent. Skills can be loaded on demand and called directly from chat using slash commands like /skill-name. Learn more: https://t.co/EZHrB20F6l

Introducing Unsloth Studio ✨ A new open-source web UI to train and run LLMs. • Run models locally on Mac, Windows, Linux • Train 500+ models 2x faster with 70% less VRAM • Supports GGUF, vision, audio, embedding models • Auto-create datasets from PDF, CSV, DOCX • Self-healing tool calling and code execution • Compare models side by side + export to GGUF GitHub: https://t.co/2kXqhhvLsb Blog and Guide: https://t.co/ENuTWal5AA Available now on Hugging Face, NVIDIA, Docker and Colab.

Ayyy @steipete at the tent! We’re at GTC park setting people up with Claws! DGX sparks available to run it locally Nemoclaw team is here answering questions. 8am-5pm today & tomorrow, til 2pm Thurs 🤙 https://t.co/zmVKKjXvUZ

Today I'm releasing my entire newsletter archive (350+ posts) and all podcast transcripts (300+ episodes) as AI-friendly Markdown files. Plus an MCP server and GitHub repo. A few months ago I shared my podcast transcripts on a whim, and y'all built the most amazing things—an RPG game, a parenting wisdom site, infographics, a Twitter bot, and 50+ other projects. Let's see what happens when I give you even more data. Grab the data here: https://t.co/xEPCcPiZHO. Paid subscribers get all of the data (some 350 posts and 300 transcripts). Free subscribers get a subset. I don’t think anyone’s ever done anything like this before, and I’m excited to give you this excuse to play with that AI tool you've been meaning to try. Here’s my challenge to you: build something, and let me know about it. I’ll pick my favorite and give you a free 1-year subscription to the newsletter. Just post a link to your project in the comments here: https://t.co/yJDmuCQfOx. If you’ve already built something, slurp in this new data and submit it, too. I’ll pick a winner on April 15th. Check out today's newsletter post for inspiration on what you could to build: https://t.co/yJDmuCQfOx LFG.

Happy St. Patrick’s Day to all who celebrate! 🍀 May your tests be as green as your contributions. Sláinte! 🟩 💚 🍻 https://t.co/I79N6TswRL

Been exploring a new way to explore AI research papers to discover deeper insights. Agents are at the center of it. So far, I've built this little interactive artifact generator in my orchestrator to visualize things. This allows me to change views and insights (on-demand) from 100s of papers. Just scratching the surface here. More to share soon.

Glad to be highlighted during the GTC keynote. We’re building the systems that bring frontier research to production and accelerate inference (and beyond). Come join us at Together! @NVIDIAAI @togethercompute https://t.co/3KeWBqBrjR https://t.co/jxIr8sMk5S https://t.co/fG7LenXBCy

Anyone else impressed nvidia got the install script on the main domain, root path? Imagine the meeting of lawyers, security and infra https://t.co/dy4itBG7JR

BREAKING: William Kelso Flournoy, a man pretending to be a woman, was arrested in WA state after he allegedly R*PED an infant girl. Why does this keep on happening?! https://t.co/iHnp9VF1gP

"Our most important consideration is recruiting the best people. The output of any company is the vector sum of the people within it." - @elonmusk https://t.co/NgetAvmzaj

Mixture-of-Depths Attention paper: https://t.co/OUgyAIQox7 https://t.co/IiQmDjq51p

+1 more

NEWS: Globe Telecom said it has successfully tested Starlink’s Mobile service in the Philippines, allowing phones to connect in areas with no signal. The pilot was done in Rizal, Batangas, and Bataan, where users were able to send messages, make calls, and use data even without nearby cell towers. "This will be our lifeline, especially during disasters and our complementary coverage in areas where terrestrial network is not available," said Joel Agustin, Senior Vice President for Service Planning and Engineering at Globe. "The service will also address the connectivity requirements of GIDA (Geographically Isolated and Disadvantaged Areas) communities and strengthen coverage across the country's territorial boundaries," he added.

I am actually hosting a live session on the topic of building and optimizing agent skills in a few days. https://t.co/1cPggXyv8N

Wow, this was unexpected. I seriously cannot thank @OpenAIDevs & @reach_vb enough for this. This really helps a lot, my expenses have been piling up. Means a lot to be recognized for my effort. Just like before, my goal with this account is to give back much more than I take. https://t.co/TAe6z3yDwV

The PokeAgent Challenge Competitive and Long-Context Learning at Scale paper: https://t.co/TrTvHiI3tC https://t.co/jhzZSPVj5Y

Grounding World Simulation Models in a Real-World Metropolis paper: https://t.co/yGrI2F67ej https://t.co/S56BAql8Ka

GPT-5.4 mini is available today in ChatGPT, Codex, and the API. Optimized for coding, computer use, multimodal understanding, and subagents. And it’s 2x faster than GPT-5 mini. https://t.co/DKh2cC5S3F https://t.co/sirArgn37L

Comet is the AI Browser for the Enterprise. Can be rolled out to thousands internally via MDM. Integrated with @CrowdStrike's Falcon for detecting suspicious links and files. Controllable for what and where agents can operate. https://t.co/n2g98Cei3o

🚨 Let me tell you what OpenAI just did to its own investor. > Microsoft gave OpenAI $13 billion.. to build them the infrastructure, gave them Azure, and made them what they are. > OpenAI just signed a deal with AWS - Microsoft's biggest competitor - to sell AI to the US government. Not through Azure.. Through Amazon. Read that again... They used Microsoft's money to build the product, then sold it through Microsoft's rival to the biggest customer on earth. But it gets worse. This is the same OpenAI that: > Started as a nonprofit to "benefit humanity" > Took billions from Microsoft > Tried to go for-profit > Got sued by their co-founder > Announced $500B Stargate - which collapsed > Launched "adult mode" after their own safety team begged them not to > And now just backdoored their biggest investor to chase government contracts Anthropic got blacklisted by the Pentagon for being a "supply chain risk." OpenAI saw that and said: "Hold my beer, we'll just use Amazon instead." In 3 years they went from "AI for humanity" to "AI for the highest bidder." When someone shows you who they are the first time, believe them. OpenAI has shown you twelve times.

AI is quietly becoming part of everyday PhD work. Graduate students now use it to draft papers, write code and search literature. The tension is obvious. These tools can accelerate research, but they may also weaken the very skills a doctorate is supposed to develop. At what point does assistance start replacing learning? https://t.co/7rl2geS4pH @nature @lindanordling @ResearchAfrica

Katherine Boyle says SpaceX’s legacy will be the thousands of companies that spawn from the school of Elon Musk: "What we've learned from the SpaceX story is that these kinds of companies can be built." "It is one of the most important companies in America. It's beloved by Americans across the country for what it's been able to achieve." "The legacy of SpaceX is going to be hundreds, if not thousands of companies—people who've gone to the school of Elon Musk, they've learned manufacturing and they've built companies, and built products in a way that they can take to new companies in manufacturing and defense, and drones and hypersonics—all of these other categories that SpaceX doesn't do." "We're incredibly excited about SpaceX. We're incredibly excited for what it's achieved for the American people and what it's going to lead to over the next 25 years." @KTmBoyle on @foxbusiness

Ask Grok in xAI Docs is actually useful https://t.co/M0LdzJhC7b

The newest model in the Mamba series is finally here 🐍 Hybrid models have become increasingly popular, raising the importance of designing the next generation of linear models. We've introduced several SSM-centric ideas to significantly increase Mamba-2's modeling capabilities without compromising on speed. The resulting Mamba-3 model has noticeable performance gains over the most popular previous linear models (such as Mamba-2 and Gated DeltaNet) at all sizes. This is the first Mamba that was student led: all credit to @aakash_lahoti @kevinyli_ @_berlinchen @caitWW9, and of course @tri_dao!