Bedram Tamang

🚨 BREAKING: NVIDIA just removed the biggest friction point in Voice AI They open-sourced PersonaPlex 7B, a real-time conversational model. It listens and speaks simultaneously to handle natural interruptions and overlaps. 100% Open Source.

Wait a sec - Google's new gemma-4-E4B is running at 400 tokens per second on my Macbook m5, while Claude Code does 90/tks? And it's free? Same-ish quality as ChatGPT 5.2 I was using last month? 🤯 #gemma #codex #claudecode

🙌 Hope you like this package! Here are all the links: Blogpost: freek.dev/3058-whats-new… Docs: spatie.be/docs/laravel-a… GitHub: github.com/spatie/laravel…

Prepping a major new release of spatie/laravel-activitylog All feedback welcome! PR: github.com/spatie/laravel…

Refactored 141 useEffect calls today after feeding my agent this. react.dev/learn/you-migh…

We are happy to announce that Vite+ is now fully open source under MIT license. Free for everyone! Go check it out today: viteplus.dev

🔗 Building Multi-Agent Workflows with the Laravel AI SDK laravel.com/blog/building-… #laravel #ai

Long awaited feature 🔥 inertiajs.com/docs/v3/the-ba…

Inertia 3 beta is here! 🚀 This is a big one. New Vite plugin, optimistic updates, standalone HTTP requests, layout props, simplified SSR, and Axios has been replaced with a built-in XHR client. Here's what's new 🧵

New course: Build and Train an LLM with JAX, built in partnership with @Google and taught by @chrisachard. JAX is the open-source library behind Google's Gemini, Veo, and other advanced models. This short course teaches you to build and train a 20-million parameter language model from scratch using JAX and its ecosystem of tools. You'll implement a complete MiniGPT-style architecture from scratch, train it, and chat with your finished model through a graphical interface. Skills you'll gain: - Learn JAX's core primitives: automatic differentiation, JIT compilation, and vectorized execution - Build a MiniGPT-style LLM using Flax/NNX, implementing embedding and transformer blocks - Load a pretrained MiniGPT model and run inference through a chat interface Come learn this important software layer for building LLMs! deeplearning.ai/short-courses/…

I hope you'll like al of the changes! 🕷️ Links: Docs: spatie.be/docs/crawler/v… Repo: github.com/spatie/crawler Blog post: freek.dev/3039-a-new-maj…

Using PostgreSQL as a Dead Letter Queue for Event-Driven Systems diljitpr.net/blog-post-post…

Having a lot of fun with @QuiverAI in my down time between baby responsibilities! So exciting to have an image model that does vector graphics, especially for the sort of work we do. quiver.ai

Really neat approach. Reflections API can be very clunky freek.dev/3030-a-clean-a…

@_vmlops Just so many bugs 🤪

Google open-sourced its internal AI agent framework ADK (Agent Development Kit) A true code-first framework to build, test, deploy & orchestrate AI agents using real software practices Supports Python, Java, TypeScript, Go, event-driven runtime, built-in tools, A2A communication & seamless integrations This isn’t experimental It’s the same system powering Google’s own AI products Repo - github.com/google/adk-pyt… The full docs - google.github.io/adk-docs/

Google just killed the document extraction industry. LangExtract: Open-source. Free. Better than $50K enterprise tools. What it does: → Extracts structured data from unstructured text → Maps EVERY entity to its exact source location → Handles 100+ page documents with high recall → Generates interactive HTML for verification → Works with Gemini, Ollama, local models What it replaces: → Regex pattern matching → Custom NER pipelines → Expensive extraction APIs → Manual data entry Define your task with a few examples. Point it at any document. Get structured, verifiable results. No fine-tuning. No complex setup. Clinical notes, legal docs, financial reports, same library. This is what open-source from Google looks like.

Go to hetzner.com Set up a $2.50/mo VPS Add your SSH key when creating Login, first install Tailscale on VPS and laptop Once it works add the Tailscale subnet (google the address) to Hetzner inbound firewall SSH port, don't add any more inbound ports Now only u can SSH into it, safe Then install OpenClaw on the VPS via SSH Don't connect any of your accounts, make a Telegram bot and add that Then play with it You have to play with it to get it

Layer order is the main issue. Every code change rebuilds node_modules. COPY package.json first, then npm install, then COPY the rest. Now deps are cached unless package.json changes. Multi-stage build for size. Build in full node image, copy only dist/ and prod deps into node:18-alpine. 2.8GB should drop probably to <200MB.

Research papers you must read for AI Engineer interviews - 1. Attention is all you need (Transformers) 2. LoRA (Low rank adaption) 3. PEFT ( Parameter Efficient Fine Tuning) 4. VIT (Vision Transformers) 5. VAE (Variational Auto Encoder) 6. GANs ( Generative Adversarial Networks) 7. BERT ( Bidirectional Encoder Representation from Transformers) 8. Diffusion Models (Stable Diffusion) 9. RAG (Retrieval Augment Generation) 10. GPT (Generative Pre-trained Transformers)

@ollama Step 3: Download the model of your choice. Make sure that you have enough hardware to support the model that you are downloading. However, Following models are recommended for this use case.