What's one local AI workflow you wish you'd discovered sooner?

There are a lot of posts about the models and benchmarks, but I am more interested in the workflows that people use. What is one workflow that really saved you time or made your local LLM more useful? It could be anything—RAG, MCP, coding agents, organizing prompt, document indexing, automation or something else entirely. What was it, and why did it make such a big difference in your day-to-day workflow? submitted by