While hacking on #calva for some stuff, I've been blown away by GitHub CoPilot. It's fucking bananas at times (imo - tho I'm sure if I used GPT style stuff more often I wouldn't be so surprised I guess). I was about to add a test for some paredit stuff that has been completely untested so far. Here's what happened: 1. I just wrote the test case: describe('Kill/Delete... and the parens, arrows etc 2. Then it suggested the body of the case based on prior cases, but almost correctly guessed the function I was testing, which was at no point in this file written at all. 3. In fact, the 'Kill/Delete' phrasing is not really used throughout the codebase at all except in the package.json Wild world. So excited to have to compete even harder haha, or at the very least work faster in languages with copilot support

How cool! Never tried copilot.