I'm struggling to understand how these tools are actually useful. The AI doesn't have the product knowledge to catch logic bugs, and basic tests should eliminate syntactic bugs/endless loops etc. Perhaps codebase indexing could help, although it still wouldn't be able to understand the "why" behind the code. Curious if others feel the same

AI code review is like asking an intern with lots of time or a serious dev with very little time to review your code and give you their first feedback: sometimes incredibly useful, sometimes not

> The AI doesn't have the product knowledge to catch logic bugs

If catching logic bugs in your solution requires context such as product knowledge, provide this information to the AI (a person working on this would also need it, so writing the info down isn't a waste of time)

> and basic tests should eliminate syntactic bugs/endless loops etc

AI is worlds better at these kinds of tasks than almost every human programmer I've met - use the tools and become proficient with them and over a short period of time going back to old basic testing methods will feel archaic

> it still wouldn't be able to understand the "why" behind the code.

Add some comments if the why needs explanation. But the silicon and electrical impulses running your code don't get the "why" either, and it works fine.