yeah thanks for digging in, this was a bit ambiguous.

so take this to mean - 1) no AI copilot in the app and 2) no training on your data nor selling of it.

our take on AI is that we should focus on building tools that help address its limitations; one of the things we're particularly keen on is building stacked diffs into reviews as a primitive, so it's easier to review a large AI-generated (or assisted) PR. (e.g., diff 1 for API changes, diff 2 for backend wireup, diff 3 for front-end changes)

i think to do that, we're going to try and hook into the subscriptions that people already have and are paying for: Claude, Codex, rather than package our own, but some of that is a bit hacky to do.

hope that's clearer and thank you for asking