this is actually a really interesting problem i've been pondering quite a bit recently. what would it actually take to get to actually secure computing assuming one was able to start with a clean slate? it seems like getting away from memory unsafe languages is a pre-requisite, but beyond that... how do you build a computer system with thousands of pieces of software authored by different people with thousands of inputs that are controlled by different entities that produces output you can really trust, and verify... like not tamper resistant, but actually tamper proof with verifiable outputs. and well, also not a huge step back in performance and usability.

in reality, maybe something like this that builds on existing primitives to get there may be the only real, feasible thing... but assuming you could start from scratch, what would it look like?