How do you find bugs using ML
TLDR;
I wasn't able to make it work, mainly due to a lack of training data and clever ideas.
What is a bug ?
Maybe one of the biggest issue that I did not realise is that the concept of bug is not something clear and well-defined. Often bugs are the result of the interaction between the specification (of the client / product owner), the designer and the developper. The most clearly defined bugs are the ones that totally break a feature. However, these bugs are so critical and so obvious that they never get merged or even committed, so they rarely appear inside the Git history and we don't have access to them as training data.
Details
Challenges Ideas related to automated bug detection