It might be better to link to https://jlevy.github.io/squares/ - it's an actual written explanation of what was achieved, and not just the git repo.

Also, specifically, this claims a new lower bound, not a smaller packing than was previously known.