I've been happily using GitHub Copilot for 7-8 months, primarily in Visual Studio and VS Code, mostly with the built-in flagship models and have felt like the output is worth the cost. Lately I've been playing with a lot of different local LLM models and decided to try using Qwen3.6 35B on some non trivial programming tasks within a repository that is 10ish years old with thousands of files and multiple languages in it. I've actually been really impressed when I use the "ask" functiona...