Initialise a python dictionary with a string?

Question

I'm trying to understand one of the most elegant solutions to a Leetcode problem (Remove Invalid Parentheses) provided by the legendary Leetcode ace StefanPochmann: https://leetcode.com/problems/remove-invalid-parentheses/discuss/75028/Short-Python-BFS

What I find confusing is that he initalises a dictionary (the level variable below) with a string (the s input variable):

def removeInvalidParentheses(self, s):
    def isvalid(s):
        ctr = 0
        for c in s:
            if c == '(':
                ctr += 1
            elif c == ')':
                ctr -= 1
                if ctr < 0:
                    return False
        return ctr == 0
    level = {s}         # what the heck?
    while True:
        valid = filter(isvalid, level)
        if valid:
            return valid
        level = {s[:i] + s[i+1:] for s in level for i in range(len(s))}

I've never seen such an idiosyncratic dictionary initialisation. Why would this make sense?

I guess I code post this question on Leetcode but experience-wise it takes quite a bit of time to get a response there - I hope to get a faster response here.