[cineticdaffodil]

Can a llm predict the price of a change to a codebase in tokens and predict the origin of the price, aka cam it see good and bad architecture?

[arjie]

Some code bases better than others but the top models can.

[mathgeek]

Ask it to do so, would love to know your results.

[cineticdaffodil]

I did, the problem is that you need a sort of standardized feature change to compare similar repos. So the metric is relative only to same projects lacking that same feature. So no, its a useless thing. No architecture score comparisson between apples and oranges.

[cineticdaffodil]

https://chatgpt.com/s/t_6a64f4a6e86c8191b41cb6a810359208

[coffeebeqn]

Sure and having some in repo documentation markdown makes it a lot easier