This task have linear time solution:

def solve(s):
    uniq = set()
    beg = 0
    end = 0
    best_beg = 0
    best_end = 0
    while end < len(s):
        ch = s[end]
        if ch in uniq:
            if best_end - best_beg < end - beg:
                best_end = end
                best_beg = beg
            while s[beg] != ch:
                uniq.remove(s[beg])
                beg += 1
            beg += 1
        else:
            uniq.add(ch)
        end += 1
    if best_end - best_beg < end - beg:
        best_end = end
        best_beg = beg
    return s[best_beg:best_end]

print(solve("abcd"))
print(solve("aba"))
print(solve("abcadbef"))
print(solve("aaaaa"))

which is based on moving across continuous series of unique characters from the beginning of string.

Compare with solution from http://blog.gainlo.co/index.php/2016/10/07/facebook-interview-longest-substring-without-repeating-characters/

def solution(s):
    beg = 0
    end = 0
    longest = ''
    uniq = set()

    while end < len(s):
        end += 1
        ch = s[end - 1]
        if ch not in uniq:
            uniq.add(ch)
            if end - beg > len(longest):
                longest = s[beg:end]
            continue

        while beg < end - 1:
            if s[beg] != ch:
                uniq.remove(s[beg])
                beg += 1
            else:
                beg +=1
                break
    return longest

(which I OCRed with tesseract and edited since they provide PNG image instead of text for code).

They unnecessary cut longest string with longest = s[beg:end] each time they expand region.

And have ugly:

while beg < end - 1:
    if s[beg] != ch:
        uniq.remove(s[beg])
        beg += 1
    else:
        beg +=1
        break

with unnecessary check beg < end - 1 because we actually look for condition s[beg] == ch. That corresponds to my:

while s[beg] != ch:
    uniq.remove(s[beg])
    beg += 1
beg += 1

so we skip up to duplication character position and move beyond it and it will be before end by guarantee of our uniq set.