[muelless] Reviewing exercise submission

[silasmue]

I had no access to Copilot in time so I used ChatGPT for the 2nd review.

My Repository: https://github.com/silasmue/heat-solver
My PR: silasmue/heat-solver#1

Code Generation using RAI

This worked suprisingly well. The AI-response was accurate and the code worked first try. The response was super fast.
In my opinions for boilerplate code and things that are already implemented many times by many people AI can as we see implement the code perfectly.

My review

See the PR for the issues I found in the AI's code.

ChatGPT review

This worked also well and was a bit more detailed. The only problem I had, it had problems separating the files I submitted, which it pointed out as problem 1. The rest of the review looked good. I think if I would have used a payed plan for example Copilote, ChatGPT+ or whatever has a code agent the result would have been even more accurate.

Challenge Reviews of the cuML project

When looking at some closed PR's of the cuML project the code reviews are very heterogenouse. For some PR's they want an implementation strategy before implementation and reviewing code in very detail. For some other PR's the code gets basically just merges with a minimal comment. For example see rapidsai/cuml#7589 for very little details, and rapidsai/cuml#7550 where is more like a discussing. I feel it is very different on who contributes and what is contributed (fix/feature).