Also it uses GPT4 by default. Seems like that could get expensive depending on how frequently it calls the API. And how much context (tokens) it sends for each prompt.
However for a professional developer $10 a day for something that's much more helpful than Copilot could be just a rounding error. Especially if that "much more helpful" translates to a significant productivity boost over and above Copilot.
Big companies aren't likely to go for something like this because of IP and confidentiality concerns. But for a solo or indie developer I could possibly see it.
Does it perform a lot worse with GPT 3.5? That could drop cost hugely if it's close to as good.
Also it remains to be seen how GitHub Copilot X will perform, whenever they release it.
End of the day, it's a reall cool experiment. Thanks for sharing!