Ask HN: What would justify writing a kernel in 2026?

If you mean an OS kernel, it would be interesting to create a fully async kernel using async/await. I've only seen one project working on this.

What project was that?

Writing an OS in Rust: Async/Await (2020) https://os.phil-opp.com/async-await/

Does writing a kernel need to be justified?

One would be if you want to design it differently than the existing kernels do.

GNU/Hurd needs a microkernel

Because you want to?