This expression calculates the position of the alphabet by subtracting 96 from its Unicode code point. Since lowercase 'a' has a Unicode code point of 97, subtracting 96 gives 1, which represents the position of 'a'. Similarly, 'b' would be 2, 'c' would be 3, and so forth.

can someone explain me why -96 please?

The problem asked to use the filter function. How did this code past the "filter in code" test?

str turns the integer into a string. It's the difference between the number 55 and the string "55". To check whether the string "5" is within the number, you first have to convert the number to a string by using str.

That's why this kata is easy, and this one is not: https://www.codewars.com/kata/621f89cc94d4e3001bb99ef4

exactly what I thought.

How can you solve it without creating an array?

In isolation using 96 is fine, but structuring it the way Kosmarik proposed is more extensible in the future (for example, if you wanted to extend the function to return a character's position relative to any other character, it'd be easier to implement that by just replacing 'a' with a function argument).

why would it be better? Every dev knows the ascii table and where to look it up in seconds and everyone who reads the solution instantly knows why to subtract exactly 96.
Your solution is much worse to read

You can do alphabet.lower()

doesn't work for uppercase.

I did this to practice using reverse (which I always confuse with reversed) to help myself differentiate them, although list slicing would have been simpler

Better than mine.