Image created with OpenAI GPT-Image-1. Image prompt: Cheesy late-night infomercial freeze-frame—tool chest overflowing floppy disks stamped “OPENSOURCE CODE-CRATE™”; orange shop-light, analog hiss, high-resolution
Anthropic just dropped a free course on building AI Apps with MCP. Learn to connect AI Agents to external data sources like GitHub, Google Docs, local files using MCP. 100% free. https://x.com/Saboo_Shubham_/status/1929916710682783915
II-Medical-8B-1706 is our latest state of the art open medical model 💡 Outperforms the latest @Google MedGemma 27b model with 70% less parameters 🤏 Quantised GGUF weights, works on <8 Gb RAM 🚀 One more step to the universal health knowledge access that everyone deserves ⚕️ https://x.com/ii_posts/status/1934959488710094990
Intelligent Internet introduced II-Medical-8B-1706, an updated version of its open medical model Capable of running on <8GB RAM, the AI outperformed Google’s MedGemma 27B across benchmarks despite 70% fewer parameters https://x.com/rowancheung/status/1935247303524114645
Today we’re announcing we’re going to build an AI cloud together with @MistralAI”” – Jensen @nvidia GTC today unveiling Mistral Compute This is a massive win for America and for open source Open models on US chips wil be the template for AI infrastructure buildouts globally https://x.com/AnjneyMidha/status/1932782589410226388
🚀 Introducing Cosmos-Predict2! Our most powerful open video foundation model for Physical AI. Cosmos-Predict2 significantly improves upon Predict1 in visual quality, prompt alignment, and motion dynamics—outperforming popular open-source video foundation models. It’s openly https://x.com/qsh_zh/status/1933024567011995865
Introducing “”Building with Llama 4.”” This short course is created with @Meta @AIatMeta, and taught by @asangani7, Director of Partner Engineering for Meta’s AI team. Meta’s new Llama 4 has added three new models and introduced the Mixture-of-Experts (MoE) architecture to its https://x.com/AndrewYNg/status/1935350552692658202
A big AI question is why, as LLMs get bigger, their values seem to increasingly converge on the same preferences, and this holds for Musk’s Grok and China’s DeepSeek, too. “These findings suggest that value systems emerge in LLMs in a meaningful sense, with broad implications” https://x.com/emollick/status/1934278025685901374
RT @NielsRogge: “”Hugging Face is basically the equivalent of Github in the era of software 2.0″” – Karpathy, 2025, colorized https://x.com/reach_vb/status/1935970251004313788
30 lines is all you need to create a MCP Agent using @GoogleDeepMind Gemini 2.5 and remote MCP servers. Here is an example using @cognition_labs Deepwiki MCP server to to chat about repositories. https://x.com/_philschmid/status/1930357626686296237
Does MCP Kill (Centralized) Vector Search? In the post-MCP world, AI agents will interact with the external world through MCP tools. There is a very valid question on whether you would want to have a centralized search index at all if agents can just directly interface with the https://x.com/jerryjliu0/status/1935473439948890177
With MCP, you can connect your agent to an entire ecosystem of tools – but how can you ensure your agent orchestrates those tools reliably? Tool rules in Letta allow you to specify constraints on your agent’s tool calling, so you can trade off autonomy and control. https://x.com/Letta_AI/status/1932836369128436042
Bring your Apple apps into the agent ecosystem. @mattt from @loopworkai demos iMCP, a native macOS app that connects services like Messages and Contacts to MCP agents using the Swift MCP SDK—yes, even for AI-powered birthday reminders. https://x.com/WorkOS/status/1930978482705572204
🤖 Introducing Motion for AI Supercharge your LLM with the new Motion MCP: • Generate CSS spring and bounce easings • Spring and easing visualisations • Documentation Plus Motion+ Rules and vibe coding guides for @framer, @v0, and @figma https://x.com/mattgperry/status/1932776126314918020
Claude Code can now connect to remote MCP servers. Pull context from your tools directly into Claude Code with no local setup required. https://x.com/AnthropicAI/status/1935367951542280239
Free Open Source Course on MCP – Built with Knowledge Graphs! I created a complete course on how to build an MCP Client and MCP Server using the Python SDK. All structured via knowledge graphs extracted from the repo itself. Let me show you what’s inside 👇 🧵 1/ https://x.com/dani_avila7/status/1930266266884870458
Had nothing better to do, so I turned a free day into a 12-hour hackathon and built an MCP server for @Core_LN in Rust. Because obviously, trusting your node to an AI is peak Bitcoin security.😃 Still rough around the edges, but it works! https://x.com/adityaa2011/status/1932824974022083031
Huge quality of life upgrade for devs If you haven’t already tried connecting some MCP servers in Claude Code, now’s the time to do so”” / X https://x.com/alexalbert__/status/1935384922493173861
Just dropped something we’ve been building: MonetizedMCP ( https://x.com/chetan_guevara/status/1932159875771215996
MorphIS is built for extensibility. From financial tools to robotic controllers, agents can plug into real-world systems using add-ons powered by the Model Context Protocol (MCP). Any service. Any industry. One interface. https://x.com/Morphisdotai/status/1932092824264352226
RT @chu_onthis: new mcp spec just dropped: 1. auth is fixed! at last! 2. elicitation now makes it possible for a server to ask an end-user…”” / X https://x.com/jeremyphoward/status/1935481114195542291
Something magical happens when you combine MCP with generative UI 🤩 https://x.com/fr0gger_/status/1930901099655176660
Thanks @TaskmasterAI another project complete! This is an excellent MCP for working on new projects with @cursor_ai or @claude_code Link below 👇 https://x.com/majidmanzarpour/status/1931421962120925242
Tired of opening old codebases and feeling like a stranger in your own repo? I built a couple of MCP servers so Claude can read my code, summarize it, and write notes straight into Obsidian. We chill like that. https://x.com/freshlimesofa/status/1931358615119786456
Want to code insanely fast and stay secure? Time to level up with AI powered secure code scanning from @snyksec available as a one click install from @cursor_ai MCP directory Instantly scan raw code, catch vulnerabilities as they’re born, and all without needing a build. https://x.com/edsim/status/1932789827935875444
ElevenLabs said its Conversational AI offering now supports Anthropic’s open Model Context Protocol (MCP) The development will enable ElevenLabs users to build voice agents that can easily connect to data from third-party apps like Salesforce, HubSpot, and Gmail”” / X https://x.com/rowancheung/status/1934881619086586251
Claude Code + Gemini Pro: Two AI coders teaming up. Yes, A redditor just dropped an MCP server that lets Claude Code and Gemini 2.5 Pro work together, and the results are seriously next-level. Claude kicks things off with planning, while Gemini adds deep reasoning and context. https://x.com/HarshaGadekar/status/1932132749923758195
Running mlx-lm locally with MCP using @huggingface’s tiny-agents is actually pretty easy and works quite well. A demo running Qwen3 4B with an MCP client for the local file system: https://x.com/awnihannun/status/1931755333011349831
MCP servers – OpenAI API https://platform.openai.com/docs/mcp
Turn your designs directly into code using @figma’s Dev Mode MCP server in @code – now available in open beta! https://x.com/code/status/1930300967695548919
RT @menloresearch: Meet Jan-nano, a 4B model that outscores DeepSeek-v3-671B using MCP. It’s built on Qwen3-4B with DAPO fine-tuning, it h…”” / X https://x.com/mervenoyann/status/1934909412117741601
Built an internal tool to connect any remotely hosted MCP server and test it against multiple agent architectures to perform evals. Just paste the MCP server URL and hit connect. Database is SQLite and all agent run logs are stored locally. https://x.com/trillhause_/status/1931525724713705589
Excited to announce that the local mistral-rs inference server now has full, built-in 🔌 MCP support – the emerging AI agent interface or ‘USB-C for AI’! You can seamlessly use your local LLM with any MCP, locally hosted or in the cloud, with automatic tool detection. Just https://x.com/ProfBuehlerMIT/status/1932743945651683466
Tracing + Evals w/o LangChain/Graph How to get the benefits of LangSmith (evals + tracing) + Studio (testing) w/o using LangChain or LangGraph? Here, we walk through the from scratch, using a non-LangChain/Graph agent as an example! 📽️: https://x.com/LangChainAI/status/1935706402896707657
RT @googleaidevs: 🩺 Get started with MedGemma, a collection of Gemma 3 variants built for medical text and image comprehension. Choose betw…”” / X https://x.com/osanseviero/status/1936096973691539652
moonshotai/Kimi-Dev-72B · Hugging Face https://huggingface.co/moonshotai/Kimi-Dev-72B
🧠 New in LlamaIndex: Custom Multi-Turn Memory for Agents We’ve just added a new example that shows how to build your own multi-turn memory implementation in LlamaIndex — perfect for agentic workflows that need control, transparency, and customization. Why it matters: 💬 Most https://x.com/llama_index/status/1932173369858003304
Outlines v1.0 is out!”” / X https://x.com/ollama/status/1935712844701442245
DAMN! Mistral just dropped an update to Mistral Small 3.2 24B – better at instruction following, less prone to repetitions AND much more robust at Function calling! 🔥 https://x.com/reach_vb/status/1936094433985826972
Day 1/5 of #MiniMaxWeek: We’re open-sourcing MiniMax-M1, our latest LLM — setting new standards in long-context reasoning. – World’s longest context window: 1M-token input, 80k-token output – State-of-the-art agentic use among open-source models – RL at unmatched efficiency: https://x.com/MiniMax__AI/status/1934637031193514237
MiniMax dropped M1, a new open-source reasoner LLM with: —World’s largest context window: 1M-token input, 80k-token output —SOTA agentic use among open-source models —2x faster training, via CISPO RL algorithm https://x.com/rowancheung/status/1934881517408325712
MiniMax-M1/MiniMax_M1_tech_report.pdf at main · MiniMax-AI/MiniMax-M1 https://github.com/MiniMax-AI/MiniMax-M1/blob/main/MiniMax_M1_tech_report.pdf
Prompt engineering UX improvement ✨ Turn your prompt into a reusable template by adding variables. Just highlight any part of the text and assign it a name to convert it into a variable! https://x.com/LangChainAI/status/1936122960089432347
RT @allhands_ai: Introducing the OpenHands CLI, a new coding CLI that: – Has top accuracy (similar to Claude Code) – Is completely open so…”” / X https://x.com/LoubnaBenAllal1/status/1935367022403367279
We’re open sourcing an entire Chatbot building platform in 3 days 👀 Crawl websites with @firecrawl_dev, create chatbots trained on multiple sources, and get auto-generated API endpoints for every bot. The perfect chatbot starter kit is coming. Stay tuned! https://x.com/nickscamara_/status/1934263691064803720
Love this contrarian view from @mmitchell_ai in @FinancialTimes today! For me, AGI/singularity/super intelligence are interesting concepts—but they feel quite abstract, vague, and easy to use for manipulation, myth-making, or self-aggrandizement. If they’re our only focus, they https://x.com/ClementDelangue/status/1935705674584924385
Microsoft-backed Mistral launches European AI cloud to compete with AWS and Azure | VentureBeat https://venturebeat.com/ai/microsoft-backed-mistral-launches-european-ai-cloud-to-compete-with-aws-and-azure/
RT @lmarena_ai: 🚨Breaking: New DeepSeek-r1 (0528) just tied for #1 in WebDev Arena, matching Claude Opus 4! More highlights: 💠 #6 Overall…”” / X https://x.com/ClementDelangue/status/1934714392693588415
🚀Exciting News! We’re thrilled to introduce OpenWBT — the open-source, whole-body VR teleoperation system designed for humanoid across various robot types & virtual/physical worlds! 🤖✨ 💡Developed through collaboration between Galbot and Tsinghua University researchers! 🔥 https://x.com/GalbotRobotics/status/1932082760736190588
RT @reach_vb: WOW! DeepMind *just* dropped Magenta Real-time – Apache 2.0 licensed 🔥 > 800M params transformer, trained on ~190K hours of…”” / X https://x.com/osanseviero/status/1936182860228034902
So excited to welcome Google’s model #1000 at Hugging Face: Magenta Real Time!🤯 🎷Music generation model ⚡️Real-time 👀Permissive license 🤏800 million parameters Model: https://x.com/osanseviero/status/1936170526931615849
6/ LlamaFS is a self-organizing file manager. It automatically renames and organizes your files based on their content and well-known conventions (e.g., time). It supports many file types , including images and audio. Super cool project by @AlexReibman! https://x.com/AtomSilverman/status/1932988830904365480
Chatterbox is a free open-source voice cloning model with emotional tone control https://the-decoder.com/chatterbox-is-a-free-open-source-voice-cloning-model-with-emotional-tone-control/
stop using VLMs blindly ✋🏻 compare different VLM outputs on a huge variety of inputs (from reasoning to OCR!) 🔥 > has support for multiple VLMs: Gemma 3, Qwen2.5VL, Llama4 > recommend us new models or inputs, we’ll add 🫡 https://x.com/mervenoyann/status/1935708014645784713
Serving Large Language Models on Huawei CloudMatrix384 – Integrates 384 Ascend 910C NPUs, interconnected via an ultra-high-bandwidth, low-latency UB network, optimized for large-scale MoE and distributed KV cache access – DeepSeek-R1 on CloudMatrix-Infer hits 2k tokens/s decode https://x.com/arankomatsuzaki/status/1934812692587663697
Groq announced it is now directly integrated as an inference provider on the Hugging Face Playground and API The move adds to the list offered by HF, giving users the flexibility to choose their preferred inference with unified access and billing https://x.com/rowancheung/status/1934881596613525793
Mistral just launched Mistral Compute — its own integrated AI infrastructure stack. It offers enterprises private access to NVIDIA GPUs, plus orchestration, model APIs, and training services A bold move to take on American and Chinese cloud giants https://x.com/rowancheung/status/1933072085288419804
If you’re running JAX and you need to grab a model checkpoint from HuggingFace, KerasHub has you covered. Load, fine-tune, quantize, export for inference.”” / X https://x.com/fchollet/status/1935045305470435484
stop writing CUDA kernels yourself we have launched Kernel Hub: easy optimized kernels for all models on @huggingface Hub 🔥 use them right away! it’s where the community populates optimized kernels 🤝 keep reading 😏 https://x.com/mervenoyann/status/1934555936561242144
🚨 @huggingface + @GroqInc = Game Changer for Builders Groq is now integrated into the Hugging Face Playground and API, unlocking the speed to bring your most ambitious ideas to life — faster than ever. From smarter agents to real-time copilots, the future is yours to https://x.com/GroqInc/status/1934696430276624820
RT @GroqInc: 🚨 @huggingface + @GroqInc = Game Changer for Builders Groq is now integrated into the Hugging Face Playground and API, un…”” / X https://x.com/mervenoyann/status/1934717979121406310
Nice improvement to the Hugging Face hub – you can now filter models with a given size range that run in mlx / mlx-lm: https://x.com/awnihannun/status/1934655784547439008
🚀 Excited to launch Qwen3 models in MLX format today! Now available in 4 quantization levels: 4bit, 6bit, 8bit, and BF16 — Optimized for MLX framework. 👉 Try it now! Huggingface： https://x.com/Alibaba_Qwen/status/1934517774635991412
The response to our first-ever Llama Startup Program was astounding, and after reviewing over 1,000 applications we’re thrilled to announce our first group. This eclectic group of early-stage startups are ready to push the boundaries of what’s possible with Llama and drive https://x.com/AIatMeta/status/1935004555210846271
Upgraded from Llama 3 to Qwen3 as my go-to model for research experiments, so I implemented qwen3 from scratch: https://x.com/rasbt/status/1936041873099063333
@MistralAI @64 Multilingual dexterity: The model excels in maintaining high-fidelity reasoning across numerous languages. Magistral is especially well-suited to reason in languages including English, French, Spanish, German, Italian, Arabic, Russian, and Simplified Chinese. https://x.com/sophiamyang/status/1932452468094312640
Mistral Small 3.2 is all you need, Apache weights btw”” / X https://x.com/qtnx_/status/1936093789442973902
RT @MistralAI: Introducing Mistral Small 3.2, a small update to Mistral Small 3.1 to improve: – Instruction following: Small 3.2 is bett…”” / X https://x.com/GuillaumeLample/status/1936104812447514968
🚀 Migrating your vector collections just got a whole lot easier We’ve released a beta, open-source CLI tool for streaming vectors between Qdrant instances (OSS ↔ Cloud), across regions, and from other vector DBs. ✅ Run live, resumable batch transfers ✅ Adjust collection https://x.com/qdrant_engine/status/1934898309878972502
Rednote releases its first open-source LLM with a Mixture-of-Experts architecture https://the-decoder.com/rednote-releases-its-first-open-source-llm-with-a-mixture-of-experts-architecture/
Yesterday we launched VibeKit 🖖 an open-source SDK to run coding agents in secure sandboxes. Today we’re excited to release VibeKit Onboard 🛹 – Let your users integrate your product by clicking a button in your docs. https://x.com/pelaseyed/status/1932907707666686171
Moonshot AI launched Kimi-Dev-72B, a new open-source coding model for software engineering tasks It achieves SOTA results on SWE-bench Verified software tasks, surpassing open-source rivals like DeepSeek R1, V3, and Devstral https://x.com/rowancheung/status/1934881573490331768
Thrilled to introduce Kimi-Dev-72B, our new open-source coding LLM for software engineering tasks. Kimi-Dev-72B achieves 60.4% resolve rate on SWE-bench Verified, setting a new SoTA result among open-source models. (1/5) https://x.com/yang_zonghan/status/1934652763985838585
🏆 MiniMax: the 2nd most intelligent open weights model after DeepSeek R1 And 1Mn token context window. The model is fully open source under an Apache 2.0 license. It is available now on Hugging Face and GitHub for unlimited commercial use. They release two versions of https://x.com/rohanpaul_ai/status/1935512689310044501
RT @iScienceLuvr: This is pretty cool, a DeepSeek researcher open-sourced “”nano-vLLM””, a lightweight implementation of vLLM in ~1,200 lines…”” / X https://x.com/jeremyphoward/status/1935994549882830993
RT @andimarafioti: 📢 A new open-source OCR model is breaking the internet: Nanonets-OCR-s! Nanonets understands context and semantic struc…”” / X https://x.com/ClementDelangue/status/1934974278287479182
RT @kyutai_labs: Kyutai Speech-To-Text is now open-source! It’s streaming, supports batched inference, and runs blazingly fast: perfect for…”” / X https://x.com/clefourrier/status/1935701954358890806
Snouzy/workout-cool: 🏋 Modern open-source fitness coaching platform. Create workout plans, track progress, and access a comprehensive exercise database. https://github.com/Snouzy/workout-cool
2/ @llama_index provides a data framework for LLM applications, facilitating efficient data indexing and retrieval. It is the simplest framework for connecting custom data sources to large language models. https://x.com/AtomSilverman/status/1932988797115117740
Meet SO-101, next-gen robot arm for all, by @huggingface 🤗 Enables smooth takeover to boost AI capabilities, faster assembly (20mn), same affordable price ($100 per arm) 🤯 Get yours today! Links in thread below 👇 https://x.com/RemiCadene/status/1916751964807057515
RT @burhansebin: Kicking off the Le Robot Hackathon Miami with amazing fireside chat with @ClementDelangue from @huggingface! @LeRobotHF…”” / X https://x.com/ClementDelangue/status/1933914952697352408
RT @PTrubey: The Los Angeles Hugging Face LeRobot hackathon starts! https://x.com/ClementDelangue/status/1933945725630431536
transformers @huggingface has a nice update for video models ⏯️ we used to handle video frames through image processors, which requires hacks around sampling & misses video specific configs now we have video processors, all models are refactored 🎉 see below how to use⤵️ https://x.com/mervenoyann/status/1933531588811501884
