“The popular AI agent “Manus” launched in China is automating about 50 tasks, and the scenario is too dystopian. It’s said to be more accurate than DeepSeek. It can simultaneously perform SNS analysis, financial transactions, research, purchasing, and more. https://x.com/thinking_panda/status/1897951585990590469
“OpenAI released new DIY tools for building custom AI agents, including: A CUA model-powered Responses API with tools for web browsing, computer use, and file management An open-source Agents SDK to orchestrate single and multi-agent systems https://x.com/rowancheung/status/1899713252637991092
“Cloudflare just dropped an AI Agent framework to build AI agents that persist state, execute tasks, browse the web, and call AI models in real-time. All of that with just a few lines of Code. 100% opensource. https://x.com/Saboo_Shubham_/status/1895891753024307590
OpenAI calls DeepSeek ‘state-controlled,’ calls for bans on ‘PRC-produced’ models | TechCrunch https://techcrunch.com/2025/03/13/openai-calls-deepseek-state-controlled-calls-for-bans-on-prc-produced-models/
DeepSeek isn’t taking VC money yet — here are 3 reasons why | TechCrunch https://techcrunch.com/2025/03/10/deepseek-isnt-taking-vc-money-yet-here-are-3-reasons-why/
“QwQ-32B evals on par with Deep Seek R1 680B but runs fast on a laptop. Delivery accepted. Here it is running nicely on a M4 Max with MLX. A snippet of its 8k token long thought process: https://x.com/awnihannun/status/1897394318434034163
“Leaderboard lore: origin story by @_lewtun , and life in thread! April 2023: @nathanhabib1011 and I were working on an internal evaluation suite at the time – when the lb went public, @Thom_Wolf contacted us to see if we could grow it and industrialize the code” / X https://x.com/clefourrier/status/1900572125238378939
CohereForAI/c4ai-command-a-03-2025 · Hugging Face https://huggingface.co/CohereForAI/c4ai-command-a-03-2025
“Learn more about Interrupt and grab your ticket here: https://x.com/LangChainAI/status/1900621522475381145
“Meet the sponsors of Interrupt 🤝 Thrilled to announce the lineup of sponsors that are making @LangChainAI’s Interrupt conference possible! A big shoutout to our presenting sponsor, @CiscoCX, and all our incredible sponsors for Interrupt: @TryArcade, @Box, @Cloudflare, https://x.com/LangChainAI/status/1900621520508219473
Open R1: Update #3 https://huggingface.co/blog/open-r1/update-3
“R1 in Q4 getting to 18t/s on a new M3 Ultra. That’s what, $9K? getting there.” / X https://x.com/teortaxesTex/status/1899480424834899993
“Also, does this make Gemma3 27B to be the best non-reasoning LLM???? https://x.com/reach_vb/status/1899734270328889367
“Pretty wild to me that Gemma3 4B is competitive with Gemma2 27B (8 months old)🤯 EXPONENTIAL TIMELINES https://x.com/reach_vb/status/1899732585699533138
“getting some fantastic results running R1 (30 tokens/s!!!) on 3x abacus + two sticks with int0 quantization” / X https://x.com/cis_female/status/1899307802423632085
“Reka open-sourced Flash 3, a 21B parameter reasoning model —Matches o1-mini on benchmarks —32k context length —Ideal for on-device deployment —Finetuned on synthetic and public datasets —Powers Nexus, Reka’s platform for creating and managing AI agents https://x.com/rowancheung/status/1899713297584173449
“The team behind Manus partnered with Alibaba’s Qwen to develop a Chinese version of its autonomous agent The collaboration will integrate Manus with Qwen’s open-source models and computing infra This comes just after Manus went viral last week! https://x.com/rowancheung/status/1899713389439377678
“You know what’s common between Hugging Face, Stability AI, Mistral, LangChain, and CrewAI? They’re all open-source AI companies driving massive breakthroughs! The next big wave isn’t just AI Agents—it’s COSS AI Agents. So at Potpie AI we decided to actively support all Agentic https://x.com/aditikothari_/status/1896949254502973671
“Personal AI Agent Mind-Map This is a Multi-agent. Tech Stack: – Python’s (Langchain and PydanticAI) – Supabase as DB for RAG and Memory I am building this for my Sem-end-Project. https://x.com/PratikKadam_/status/1896565808937189630
“Want to build your own @v0 or AI agent? Three open-source templates you should check out: 1. Have an AI create a deployment and users “claim” it 2. AI chatbot with code generation / execution 3. Slackbot with tool calling https://x.com/leerob/status/1897753503839236282
“Today @cohere is very excited to introduce Command A, our new model succeeding Command R+. Command A is an open-weights 111B parameter model with a 256k context window focused on delivering great performance across agentic, multilingual, and coding usecases. 🧵 https://x.com/aidangomez/status/1900169306987524440
“Alibaba’s QwQ-32B, a 32.5-billion-parameter language model, demonstrates reasoning abilities comparable to much larger models like DeepSeek-R1. Fine-tuned with reinforcement learning in two stages, it excels in math, coding, and other forms of problem-solving while remaining” / X https://x.com/DeepLearningAI/status/1900351166086537659
“Alibaba’s Qwen team dropped QwQ-32B, a reasoning AI that matches or surpasses DeepSeek-R1 at much less cost —20x smaller than DeepSeek-R1 —Priced $0.20 per million input and output tokens —Open-sourced under Apache 2.0 Here it is running on M4 Max: https://x.com/rowancheung/status/1897554323489325517
“Llama models were used to develop India’s first open source audio language model — Shuka v1 ➡️ https://x.com/AIatMeta/status/1900271556917682405
“Folks, we have set up a github repo for QwQ, specifically providing evaluation scripts for you to easily test the benchmark performance of reasoning models, and also reproduce our reported results. We provide step-by-step guidance for you to run the evaluation, and we hope this” / X https://x.com/Alibaba_Qwen/status/1900595120053047452
“DeepSeek made it abundantly and nakedly clear that the models have been commoditized. If there was any doubt before, that doubt is now over. So, what is the moat right now? For me, I love Hamilton Helmer’s “Seven Powers.” If marketing is the art of decommoditizing your product, https://x.com/JonathanRoss321/status/1899169648395833703
“Many recent frontier LLMs like Grok-3 and DeepSeek-R1 use a Mixture-of-Experts (MoE) architecture. To understand how it works, let’s pretrain an MoE-based LLM from scratch in PyTorch… nanoMoE is a simple (~500 lines of code) but functional implementation of a mid-sized MoE https://x.com/cwolferesearch/status/1899172553626685532
“👋 Introducing the Enhanced Qwen Chat We are pleased to announce the latest update to Qwen Chat, designed to deliver a seamless, versatile, and user-centric experience. Explore the key features below and visit https://x.com/Alibaba_Qwen/status/1899497336889659775
“Thanks to #DeepSeek, even @CNBC [9] now talks about AI distillation, published in 1991 [1][2]. I called it “collapsing,” back then, not “distilling.” See also [9][10]. REFERENCES [1] J. Schmidhuber (JS, 1991). Neural sequence chunkers. TechReport FKI-148-91, Tech Univ. Munich, https://x.com/SchmidhuberAI/status/1899475671958929453
“@danielhanchen Hmmm seems its worse on most benchmarks than mistral 24b but 4x the context length and vision. Hard to decide which is better!” / X https://x.com/Teknium1/status/1899744944669315260
“Hugging Face Inference providers now supports Cerebras Cerebras Inference runs the industry’s most popular models at more than 2,000 tokens/s – 70x faster than leading GPU solutions. Cerebras Inference models including Llama 3.3 70B, will be available to HuggingFace developers, https://x.com/_akhaliq/status/1899502216420942232
“if you want an offline vision LM on your phone, we recently shipped HuggingSnap on test flight @pcuenq and @cyrilzakka built it with love ❤️ we would like to hear feedbacks and develop further if you like it 🥹 https://x.com/mervenoyann/status/1900492593810546774
