The coolest GitHub hack I learned this week: 1) open any repo 2) replace .com with .dev 3) browse the repository code in a VS Code instance! 🤯🤯

Meta launched new LLMs with vision capabilities (Llama 3.2 11B and 90B instead of previous 8B and 70B), smaller models (1B and 3B) for mobile devices, cheaper Quest 3S and a new product (like a dev kit), Orion (smart advanced glasses). Watch the summery youtube.com/watch?v=l_QruJ…

Meta realised vision models: Llama 3.2 11B and 90B, but with restrictions for vision in the EU. Something should be made. We can't afford to fall behind with AI, come on regulators!

I find it so interesting (and smart) that Meta / LLaMA is eliminating the dependence of their models on the HuggingFace stack. The LLaMA models now: - Have their own website to download weights. - Have one of the best LLM cookbooks that's available. - Provide extensive

Very useful for LLM Prompting. This will pack your entire repository into a single, LLM-friendly file with one command. Needed for feeding your entire codebase to Claude/ChatGPT etc. To pack a specific directory just run `repopack path/to/directory`

Confused about different embedding types? You're not alone. The landscape can be super overwhelming, so here’s a quick rundown of 5 different embedding types for your AI applications: • Dense Embeddings: The standard form, representing meaning through long sequences of

Chinese startup 01 .ai trains competitive LLM using 95% fewer resources through innovative engineering optimization. 01 .ai trained a GPT-4 competitor using just 2,000 GPUs and $3M, while achieving competitive performance. Through innovative engineering and optimization

Wait - is the new GPT-4o a smaller and less intelligent model? We have completed running our independent evals on OpenAI’s GPT-4o release yesterday and are consistently measuring materially lower eval scores than the August release of GPT-4o. GPT-4o (Nov) vs GPT-4o (Aug): ➤

Wow Flux 1.1 Pro Ultra Raw is wild. This is 100% AI 10 examples (motion with Hailuo): 1. IMG_3651.PNG casual selfie

ML Engineer vs MLOps Engineer. Do you know the difference? medium.com/@tenyks_blogge…

Embeddings on steroids: 15x smaller, 500x faster, zero effort. Vector search on top of millions of docs in seconds. no pre-indexing - reducing model size by 15x and making the models up to 500x faster Model2Vec is in Github now. Its an embedding powerhouse that distils good

OpenAI Deep Research is insane. But you don't have to pay $200/month for ChatGPT Pro subscription. 5 open source alternatives:

What is SGLang and why does it matter? SGLang is an open-source LLM inference engine that achieves up to 2-5x higher throughput than competitive solutions! And aws the first to implement multi-token prediction for DeepSeek R1 for a 1.76x speedup! 👀 TL;DR: 💡Leverages

Sam Altman Advocates for AI Proficiency as the New Essential Skill In a recent interview with Stratechery's Ben Thompson, OpenAI CEO Sam Altman emphasized the importance of mastering AI tools, likening it to the previous generation's focus on learning to code. He highlighted

Noticing myself adopting a certain rhythm in AI-assisted coding (i.e. code I actually and professionally care about, contrast to vibe code). 1. Stuff everything relevant into context (this can take a while in big projects. If the project is small enough just stuff everything

This is bigger than you think DeepMind has unveiled an AI system called AlphaEvolve that uses Gemini models and evolutionary algorithms to independently find new ways to solve complex computational tasks - such as matrix multiplication, a cornerstone of any modern AI. Why is

Introducing Claude Opus 4 and Claude Sonnet 4. Our best models yet. The Claude 4 family is here.

OpenAI introduces GDPval—a new metric that evaluates AI models based on economically relevant tasks from 44 professions, rather than just abstract tests. Initial results show that modern models often match the quality of experts and can complete such tasks ~100× faster and more

Google Colab is officially coming to Visual Studio Code! ⚡️ You can now connect VS Code notebooks directly to Colaboratory runtimes. Get the best of both worlds: the editor you love, powered by the compute (GPUs/TPUs) you need. → goo.gle/47QTmnB

🚀 Qwen3.5-Omni is here! Scaling up to a native omni-modal AGI. Meet the next generation of Qwen, designed for native text, image, audio, and video understanding, with major advances in both intelligence and real-time interaction. A standout feature: 'Audio-Visual Vibe Coding'.