We also use it for all devs but I'm thinking of cancelling. The bugs it introduced that were too easy to glance over cost us more time in total than it really saves us imho.
Yeah my guess is that $120 million revenue is likely breakeven at this point, probably being run over by salaries and even more compute / GPU / hardware costs to run this expensive LLM service.
I've been a paying user for a while now and the chat feature has become slower and slower. It takes up to a minute now for it to reply which is too long for a productive back and forth.
Since I’ve cancelled my GHCoPilot subscription, I do find myself waiting for the autocomplete to pop up! Realizing how much of my brain I outsourced to this “feature”.
I now prefer co-developing with Azure/OpenAI because I can learn as I go. To me, with GHCoPilot, I had to know enough of the language I was working with to trust the code it produced.
Now I can ask “Break down these functions into the smallest unit/explain XYZ to be as if I was a novice” and SO many other questions.
To me, working with GPT4 is way better than working with CoPilot.
it probably wasn't a full throated lie. there are so many ways to mislead without lying when it comes to financials. like quietly lumping in your MS office subscriptions into your Azure web services numbers.