I just ran this on a personal repo I've been working on for two years. It's 250k loc of Python, that for the last year has been written with agents.

I told Claude to look at this article and the original research paper and calculate the numbers. Verbosity came out to 0.08 (lower than human, that's weird), erosion to 0.495, so between humans and agents. So according to these metrics, I have managed to contain the complexity of code.

are you able to share what you're working on that's 250k LoC?