In JavaScript, the biggest test case in the full test suite has 11000000 tiles, so it must be a linear (O(n)) algorithm. (other languages should have similar sizes)

Hi @Dmillenaar and welcome to Codewars;

Since you are new I just wanted to clarify something, in case that's what is causing the difference between your IDE and Codewars:

When you click "Test" on any kata, your code is tested against the (small, few in number) tests visible in the bottom right of the kata environment. In this kata, these are the SAMPLES that are approximately 6-7 in number, with small input strings. These are designed to be "easy" tests, to check your code logic etc.

When you click "Attempt" you will then encounter the full set of kata tests which can be very large.

If you are only copying the "small tests" into your IDE, then that is why you are timing out here and not in your IDE.

If you want to recreate the size of the full kata tests to see the performance of your code in your own IDE, then randomly generate a string of length 5000000 consisting of >, <, and . and see how long it takes on your PC.

hmm i think your solution is not O(n) thats why its failing the big tests...

Yeah, generating is being a weird bottleneck (somehow using numpy is slower in this environment?!?), so I changed the random tests to make its time similar to the time i got from JS, so maybe now you can? It depends on how good is the performance of your solution.