Blog
Notes from the farm.
What actually worked, what melted four Macs, and the exact job files behind the clips.
AI Video Infrastructure
🧠 📊 🔥
AI Video
CUDA Performance on Macs for AI Video: What MLX Actually Gives You
Unified memory beats VRAM for model size, loses on raw throughput, and lies to every free-memory API. How we priced jobs instead of guessing.
29 July 2026🔗 🚫 📉
Infrastructure
LTX Meets NVLink: Why You Can't Split One Render Across Macs
Tensor-parallel diffusion needs ~900 GB/s. Your ethernet does 0.125. Here's the maths, and the parallelism that actually works on Macs.
29 July 2026🌙 🤖 🌅
Infrastructure
Autopilot: Queue the Night, Read the Report
An unattended render farm needs four rules and a pause button. Here's the autopilot that runs our Macs overnight — and what it refuses to do.
29 July 2026🖥️ 🔌 ⚡
AI Video
Who Needs CUDA? 4 Macs, One Folder, 4× Video Throughput
We turned four M4 Macs into an AI video render farm with a $20 switch and one shared folder. No NVIDIA, no cloud bill, near-linear scaling.
29 July 2026