I'm a pseudonymous indie builder running local AI on a 16GB Mac mini M4. Bought the machine on April 23, 2026 and started pulling models. One month later there are 13 daemons running on it every day — job application grading, hobby research with web grounding, a daily briefing generator, Korean phrase drills with audio, a wiki ingest pipeline that turns PDFs into a searchable knowledge graph, and a few other things.

No NVIDIA cards. No Mac Studio. No cloud GPU bill. Just the Mini, Ollama, llama.cpp for the heavy model, and Claude as a development partner the whole way.

This site is where I write down what worked, what didn't, and what I wish I knew on day one. Mostly: real A/B test results across models, the env vars and tricks that actually mattered, and the daemon patterns I keep reaching for.

The product

If you want the whole stack — schema, A/B verdicts, weekly workflow, optimization tips, journey log, roadmap — I packaged it as a Notion template you can duplicate into your own workspace:

The 16GB Mac Mini Local AI Setup pay what you want on Gumroad ($3 minimum). Living template, free updates as I keep building.

Contact

Email hello@minillm.dev. Found a bug, want a feature, or spotted something I should've done differently? Send it over, your input shapes v1.1.