MBW Extra: Leo's M4 Mac Mini Running a Local AI
Summary
Leo Laporte discusses repurposing his M4 Pro Mac Mini as a local AI inference server using Ollama, highlighting the challenges of running macOS headless. He and Christina Warren compare local models like Qwen and Gemma against cloud options, debating privacy and performance. The conversation also touches on using AI agents for personal finance management and the value of high-RAM Macs for local LLMs.
Topics discussed
Anecdote: $6 cupcake in Japan and existential philosophy
Show intro, sponsor read, and host introductions
Repurposing Mac Mini as a local AI inference machine
Technical challenges: Spotlight indexing and headless setup
Discussion on macOS limitations for server use
Troubleshooting Spotlight and SIP on Golden Gate Beta
Review of Qwen 3.8 and Kimi K3 models
Using AI agents for personal finance management
Local tracking tools and NAS backup strategies
Running various local models like GLM and Deep Seek
Hardware regrets: RAM capacity and Mac Studio vs Mini
Value assessment of the Mac Mini and closing remarks
Listen ad-free on Castria