I sure hope so, and given Copilot currently integrates with VSCode and its amazing language server APIs, I imagine this could be a (relatively) simple win to ma

I sure hope so, and given Copilot currently integrates with VSCode and its amazing language server APIs, I imagine this could be a (relatively) simple win to make a significant bump in model accuracy. I do wonder if Copilot has any idea of the structure of what it's generating though. Does it know when something is a comment, a variable, or a piece of syntax?