Want to build an AI Agent?
I made a free cookbook for creating your own news research agent with open-weight GPT-OSS models — no GPU, no setup.
Searches news → pulls articles → summarizes w/ sources → runs in a Gradio chat UI.
Check it out: huggingface.co/spaces/fdaud...
GPT-OSS:
- 5M downloads in <1 week on @huggingface 🚀
- 400 new models
- lready outpacing DeepSeek R1’s launch numbers, and that’s without counting inference calls
- also the most-liked release of any major LLM this summer
You can run gpt-oss-20B on Google Colab 🤯
Well, it took just 2 hours for GPT-OSS to hit #1 on @hf.co. Don’t remember seeing anything rise that fast!
Demo: huggingface.co/spaces/fdaud...
Blog post from the team working on this: huggingface.co/blog/manu/co...
AudioRAG is becoming real! Just built a demo with ColQwen-Omni that does semantic search on raw audio, no transcription needed.
What’s exciting: it skips transcription, making it faster and better at capturing emotion, ambient sound, and tone, surfacing results text search would miss.
⚠️ AI isn’t just a tool—it’s also supercharging medical misinformation. From fake studies to hallucinated citations, the risks are real when governments and charlatans lean on unchecked AI. www.theguardian.com/commentisfre...
🤖 On the farm, AI is running the show: self-driving tractors, robot fruit pickers, and drones are making “hands-off” farming real. The future of food? More tech, less sweat. www.wsj.com/tech/autonom...
🧠 Nearly 3/4 of teens are turning to AI companions for advice, friendship, and comfort. The upside? New connections. The risk? Losing touch with real-world relationships. Wild times for growing up. www.cnn.com/2025/07/16/h...
💰 Ex-OpenAI stars just raised a jaw-dropping $2B for Thinking Machines Lab. The AI talent wars are heating up, and the next breakthrough might come from this stealthy new contender. www.wired.com/story/thinki...
🤝 Nvidia’s CEO Jensen Huang is walking a tightrope in Beijing, balancing US-China tech rivalry while keeping Nvidia at the heart of the AI revolution. The stakes? Trillions and global influence. www.nytimes.com/2025/07/16/b...
Today’s reads: Nvidia’s CEO navigating US-China tensions, a $2B AI upstart, and why so many teens are confiding in chatbots.
Here’s what I found most interesting 👇
The best part? It's only a 2.2B model! huggingface.co/HuggingFaceT...
more info here: huggingface.co/blog/smolvlm2
and more modelshuggingface.co/models
You might not have heard of Moonshot AI, but within 24 hours, their Kimi K2 model shot to the top of the Hugging Face trending models. So… who are they, and why does this matter?
huggingface.co/blog/fdauden...
AI-powered browsers like Perplexity’s Comet promise to do your web surfing for you. But do they really save time, or just add more noise? 🌐 www.nbcnews.com/tech/tech-ne...
Google just inked a $3B deal for hydro power to run its AI data centers. Big Tech is scrambling for clean, reliable energy as AI’s appetite explodes. ⚡ www.ft.com/content/d8be...
Cognition (the Devin AI agent crew) snapped up Windsurf, a fast-growing AI coding startup. The battle for AI developer tools is getting fierce. 💻 techcrunch.com/2025/07/14/c...
Nvidia just got the OK to sell AI chips to China after its CEO met Trump. Tech, trade, and geopolitics: all on the table. www.wsj.com/tech/nvidia-...
Meta’s superintelligence lab is debating ditching open-source AI for a closed model www.nytimes.com/2025/07/14/t...
+ is dropping hundreds of billions on giant AI data centers www.bbc.com/news/article...
Today’s headlines are all about AI power plays: Meta’s open-source future is on the line, Big Tech is pouring billions into data centers and energy, and the AI talent wars just hit a new level.
Let’s dive into the 5 most important stories of the day 👇
Kimi K2 just hit #1 on @hf.co trending models in <24 hours!
This MoE powerhouse packs 1T params with 32B active - crushing coding challenges and autonomous agent tasks.
🧑💻 Want to try it? SmolLM3 is fully open-source, with all training recipes and code released. Jump in, run it locally, and see how far a smol model can go! huggingface.co/blog/smollm3
🤖 Dual-mode genius: Switch between fast answers or deep reasoning with simple flags (/think or /no_think). Plus, it supports tool calling for agentic workflows.
🌍 Multilingual: SmolLM3 supports 6 languages (English, French, Spanish, German, Italian, Portuguese) and shines across knowledge, reasoning, math, and coding benchmarks. Small but mighty!
🛠️ Under the hood: SmolLM3 uses a transformer decoder with Grouped Query Attention, NoPE for long context, and clever training tweaks for rock-solid stability—handling up to 128k context length with ease!
SmolLM3 outperforms other 3B models and even competing with 4B giants, it’s open, efficient, and ready for your toughest reasoning tasks.
Get architecture details, data mixtures, and the whole three-stage pretraining process. No secrets, no reverse engineering needed—just the complete recipe for hybrid reasoning.
🚀 Meet SmolLM3: a 3B parameter language model that punches above its weight, and comes with the *full* engineering blueprint!
