my stuff

These are either finished, in maintenance, or are usable but in active development.

dustsort

rust, theoretical cs, algorithm engineering

git repo

I designed a sorting algorithm in the block merge sort family using almost no extra memory; there’s a variant of these being used in the Zig standard library. It started as a theoretical CS project, being the first such sort to use a near-optimal 𝑛lg𝑛+𝑂(𝑛) comparisons, giving it a reliable worst-case. The overhead of my method was low, so I wrote a drop-in sort for the Rust standard library that’s safe, correct, and uses novel techniques to trounce the Rust sorts on some real-world patterns (despite their loose constraints). It’s very reliable since it has provably excellent worst-case guarantees, always sorts stably, and makes no heap allocations.

TETQ

rust, research tools, data structures

git repo

I ported a high-performance algorithm and its backing data structure from my private project to a public CLI tool. It accelerates competitive research fields in modern Tetris like creating and analyzing opening techniques. It was used in the discovery of various techniques by significantly reducing latency between finding a candidate and computing its probabilities. I wrote about the algorithm in a blog post.

Butterfly Merge Sort

c, high performance algorithms

git repo

It’s a cool experiment that I did to show that a generic (over types and strict weak orderings) in-place mergesort can be really fast in practice. It targets modern CPU architecture to compete with SoTA quicksorts under the same rules, and somehow can beat them on random data. To the best of my knowledge, there isn’t even an out-of-place (using a lot of extra space) generic mergesort that does that. I wrote about it in a blog post too.


@2026 bzy.