Random tests sometimes times out on its own.

Description updated.

Fixed.

I really don't want to play with input types and all that stuff to make it more fluffy (in some ways) at least at this point.

Thanks for feedback.

The Kata description is not perfect, I agree with it. But I'm not sure which approach is better for description (add more visual info / describe the steps in details)?! Anyway the Kata need more views and tests, etc. to minimize the changes and maximize the quality.

The meaning of n and m looks ambigious (only deducible from pictures and tests). To make matters worse, the coordinates are given in the opposite order (m axis, n axis). This should be made more clear in decription.

I'm not a big fan of this term 'incorrect sensors'. Why not just 'untouched sensors' or something similar?

Why is the return type a string? It seems to add nothing to the task itself. Wouldn't it be better if it was just a vector or array?

IDK what's best in C++, I'm just asking questions and giving suggestions :P

Wouldn't it be better to use something like vector of pairs or vector of tuples for coordinates (instead of vector of vectors of 2 elements)?

Incorrect mean that the laser will never intersect with such ones.

I will clarify it in description too.

What is an 'incorrect sensor' ?

Your solution fails on cases where parens not sorted in the usual way.

Ex: [')()()()(']
['())(']

So it's not enough to find equal amount of parens.

Not really a suggestion... the tests are fine.