The Personal AI Stack

It’s interesting to see the proliferation of “share your local LLM setup” threads—I learned a lot recently. A year ago, the conversation was almost exclusively about which centralized API to use. Now, it’s about what hardware you’re running, which models you’ve quantized, and what software you use to manage it all.

It feels like the classic unbundling moment that follows any major platform shift. First, we all used the same handful of mainframes via terminals (API keys). Now, everyone is building their own personal computer, speccing out the GPU, the OS (Ollama, etc.), and the applications (the models themselves).

This isn't just about privacy or saving on API costs. It's about control, customization, and ownership. It’s a quiet but significant shift from being a consumer of AI to being a sysadmin for your own.