Rust is great and all, but we can all agree that the compiler is pretty slow. There are a lot of talented people working on improving the situation, but I think the biggest speed improvements are hiding in radically different compiler architectures, and that's essentially impossible for 'rustc'. It's simply too big, and there are too many people working on it simultaneously for any sort of large-scale rewrite to be possible. However, somebody should still try out new compiler architecture ideas—otherwise we'll never know if they're any good.
This is a pretty massive undertaking, and I'm unlikely to finish it. But I'm hoping to take away some lessons about compiler architecture, and I hope this will inform a different compiler project that I'm considering with some friends. My goal is to write about the development experience in relatively short-form updates, and to avoid getting bogged down in the implementation too much.