10h ago · 8 min read · When porting CUDA-oriented TTS models to Apple Silicon, one subtle but important issue is BF16 behavior on Metal. This has shown up in DiaTTS work on MPS, and the same class of issue appears relevant
Join discussionMay 11 · 16 min read · To set the stage, this post covers follow-up work that made MetalGraph feel less like it is doomed to be only a pure investigation tool and more like something that could be useful in practice. MetalG
Join discussionApr 23 · 8 min read · I Run a 40GB AI Model on a MacBook. Three Months of MLX on M1 Max Has Changed How I Think About Apple Silicon. It's Just a Laptop. But It's Running a 40GB Model Right Now. I'm drafting this on a MacBook Pro. Qwen 3.6 35B-A3B MoE Q8 — about 40GB of we...
Join discussionApr 23 · 5 min read · ▶ Watch the race on YouTube: https://www.youtube.com/watch?v=2KeTDDodE0A April 22, 2026. Anthropic's Claude Code Max plan jumped to $100 a month. I ran a live three-way AI race on the exact same prompt — Gemma 31B local, Llama 70B local, and Claude...
Join discussionApr 6 · 5 min read · My M4 Max was decoding Qwen3.5 at 58 tokens per second yesterday. Today it's doing 112. Same model, same hardware, same prompt. The only thing that changed was a single environment variable. Ollama 0.19 shipped on March 31, 2026 with a preview of its...
Join discussion
Mar 18 · 3 min read · Apple Silicon changed what is possible for local AI. The unified memory architecture means ML models can run on the GPU without copying data between CPU and GPU memory. For a desktop agent that needs to process screen content in real-time, this matte...
Join discussionMar 14 · 8 min read · It started with a tweet. Google Devs posted a demo of FunctionGemma running a game, and I watched this tiny model parse natural language into structured function calls in real time. My immediate thoug
Join discussion
