For the longest time I've banged on about how useful it is write meaningful commit messages. It's something that has helped me countless times as I've had to look back on applications that are several years old, wondering why on earth I did something a certain way. Imagine now being able to take all that meaning, cram it into a LLM, and get even more insight! Find the highest risk bits of code just by analysing the sentiment of the messages? What else?