> And that's not even getting into the "waaa?" from `du -hs /nix` although I am open to that being a misleading number due to hardlinks and other trickery that

> And that's not even getting into the "waaa?" from `du -hs /nix` although I am open to that being a misleading number due to hardlinks and other trickery that du may not correctly surface Nope, du counts hardlinks correctly when it encounters the same inode multiple times in the course of fulfilling a single invocation. per https://ss64.com/osx/du.html https://ss64.com/osx/du.html : > Files having multiple hard links are counted (and displayed) a single time per du execution. Nix's disk usage profile is pretty similar to Flatpak's, or to a collection of closely related Docker containers. The difference between no Nix install and having one isn't huge. But your first installed Nix package will pull in very low level common dependencies— on Linux, everything between whatever application and the kernel (and on macOS, a bit less). Your next application will come with a smaller increase. Once you have a handful of programs installed, you no longer have big downloads for individual additions. When you have a lot installed, the difference isn't that huge, proportionally. Over time, your Nixpkgs version will rotate and you can end up with deps from old versions of Nixpkgs, which can take up a lot of extra space. But that's easy enough to manage by pinning Nixpkgs. If you ever uninstall programs, Homebrew's broken uninstallation functionality can very quickly make a Homebrew installation much (up to several times) larger than the equivalent Nix one once you have more than a handful of packages installed.