This is way older than ChatGPT, probably it was in its learning database.

I haven't been here for a while, but the point was exactly some clever code at least back then. I wonder if that shifted, or simply new people came.

It's linear, so it scales ok. Sure it's not optimized for very different lengths case, but optimized != best practice.

Best practice is right compromise between optimization and code clarity.

You should really look into regular expressions more:

function prepareNumber(n) {
    return n.replace(/^0+/, '')
}

Don't need a list around str. Strings are iterable.

Don't need to create a list, just:

sum(int(i) for i in str(n))

This was my initial solution in python, but I feel that tracking both left and right is clearer.

NaN propagation trick is fucking neat.

You don't need to store all the sums, you can update them on the fly - http://www.codewars.com/kata/reviews/54d9d8b15ca355f10100018c/groups/54ea9c0e87ed7fc0c9000a6e

This has quadratic complexity while linear is possible.

Use .each.with_index?

This has quadratic complexity while linear is possible http://www.codewars.com/kata/reviews/54d3c3bec7f9887be600019c/groups/54ea9933bc1a2557180001c7