A small update, but it really takes us quite a long time as there are so many models and we even need Mac Studio for the testing. Hope that they are helpful for MLX users!

two subagents in parallel, wwohooo

Today, we’re thrilled to unveil the Arcee.ai Foundation Models, a new family of GenAI models designed from the ground up for enterprise reality. The first release—AFM-4.5B—is a 4.5-billion-parameter frontier model that delivers excellent accuracy, strict compliance, and very

i just reached new level on Claude Code, 5 subagents working at the same time :D :D :D

$200 for Claude Max suddenly becomes too cheap

Haven’t watched a 1hr+ YouTube video in years. Got recommended a history lecture by some random Chinese high school prof — bro read the US like a textbook. Every move predicted, a year ago. Should’ve called his channel “Generative History” 😂

🥳🚀 1st installment of Mac Studios M4 Max for the “HPC+AI Synergies” teaching lab has arrived. 🚀🚀 Benefits: 👉🏻Ultra low energy footprint 👉🏻 Quiet student-friendly environment for hands on experimentation w/ parallel connectivity topologies 👉🏻Students get trained on machines

Gemini CLI is here! The power of Gemini, directly in your terminal ⭐️ 1k requests per day, free-of-charge and open source with Apache 2.0 license, with lots of tools + MCP support + so much more🔥 GitHub: github.com/google-gemini/… Blog: blog.google/technology/dev…

no so surprised, google photos back in 201x was basically the main source for their JFT-300M dataset

ByteDance just dropped their Trae IDE agent, they basically open-sourced Cursor. Everyone can made their own VSCode AI wrapper now github.com/bytedance/TRAE…

Disappointed with Cursor’s new pricing & RAM usage, so I built my own “Cursor” setup: - Ghostty terminal, 4 panes: neovim (lazyvim), Claude code, terminal, ccusage - minuet.nvim for Ollama tab completion with Qwen2.5-coder (can replace with supermaven.nvim)

The latest MLX has a CUDA back-end! To get started: pip install "mlx[cuda]" With the same codebase you can develop locally, run your model on Apple silicon, or in the cloud on Nvidia GPUs. MLX is designed around Apple silicon - which has a unified memory architecture. It uses

OHHHH HELL YEAH

i beg u pls dont Anthropic reddit.com/r/ClaudeAI/s/g…

with Anthropic max plan nerf, best way is to use it with OpenCode and a mix of GLM, Kimi, Qwen

the 5% is total people who subscribed to Max, not Claude Code Max users only