I’ve just wrapped up a three-part deep-dive series on Bloom Filters and their modern cousins. If you're curious about data structures for fast membership checks, you might find it useful.

Approximate membership query (AMQ) filters don’t tell you exactly what's in a set, but they tell you what’s definitely not there and do it using very little memory. As for me, that’s a killer feature for systems that want to avoid unnecessarily hitting the bigger persistent cache, disk, or network.

Think of them as cheap pre-caches: a small test before the real lookup that helps skip unnecessary work.

Here's what the series covers:

Classic Bloom Filter
I walk through how they work, their false positive guarantees, and why deleting elements is dangerous. It includes an interactive playground to try out inserts and lookups in real time, also calculating parameters for your custom configuration.

Counting Bloom Filter and d-left variant
This is an upgrade that lets you delete elements (with counters instead of bits), but it comes at the cost of increased memory and a few gotchas if you’re not careful.

Cuckoo Filter
This is a modern alternative that supports deletion, lower false positives, and often better space efficiency. The most interesting part is the witty use of XOR to get two bucket choices with minimal metadata. And they are practically a solid replacement for classic Bloom Filters.

I aim to clarify the internals without deepening into formal proofs, more intuition, diagrams, and some practical notes, at least from my experience.

If you’re building distributed systems, databases, cache layers, or just enjoy clever data structures, I think you'll like this one.