2026-04-19

Welcome to /lock

This is the opening note for the /lock section — the place where local-first LLM study cases and tech stack experiments live.

What lives here

Local inference setups (Ollama, llama.cpp, MLX)
Agent patterns that never touch the cloud
Offline RAG, on-device embeddings, zero-data-leak workflows
Stack deep-dives: GPU, VRAM, quantization tradeoffs

Why local-first

You own the weights. You own the runtime. Your prompts never leave the machine. For high-stakes work — legal review, medical notes, confidential code — that's not a nice-to-have, it's the whole point.