Local LLMs are actually good now, and I wasted months not realizing it
…it was trained on knowledge distilled from a much larger 397B model - so you're getting more capability than the parameter count suggests. And I also have Brave Search MCP hooked up…