This is cool! I am more interested in how you guys generated next edit training data from repos, seems like there are lots of caveats here. Would love your insights

Again amazing work! waiting for what you guys cook next