You may want to reread the article. One of his main points is how his custom algorithm accesses the dataset sequentially and thereby massively outperforms bloom

You may want to reread the article. One of his main points is how his custom algorithm accesses the dataset sequentially and thereby massively outperforms bloom filters which need random access.