def insert(x):
for hash in hashes:
bloom_filter[hash(x)] = true
The downside is that your false positive rate will increase as you insert elements. But, on the other hand, you get the guarantee of no false negatives.
This cuckoo filter has no false negatives or false positives, but their Insert() function clearly can hit a case where it returns Failure.
Bloom filters effectively represent the infinite set when it's all ones. You can always add more, and everything will return as probably in there already.
How is this different from a bloom filter?