I disagree - by and large, we don't need to "rewrite for tomorrow". Almost every significant problem in Computer Science and Software Engineering was solved (well) decades ago, often in the 1960s. Security is a bigger issue now, but it was largely solved years ago.

The problem is that we do engage in so much "rewriting", instead of leveraging known good, quality code (or at least fully-fleshed out algos, etc.) in our "new, modern, elegant, and trendy" software edifices of crap.

To me, this may be the one really good thing to come of the cloud (as opposed to the re-mainframe-ication of IT): the "OS" of the 21st century, allowing plumbing together proven scalable and reliable cloud/network services to build better software. (Again, not a new idea, this was behind the "Unix Philosophy" of pipes, filters, and making each program do one thing well. Eventually, it will be in fashion to realize this really is a better way...)

We need smaller, better software, not the latest trendy languages, insanely complex platforms that no one understands, and toolchains of staggering complexity that produce crap code so bloated that it requires computers thousands of times faster than the Crays of the 1990s just to do ordinary desktop stuff. (Seriously, folks, the Raspberry Pi 4 on the next table is a rough match for the Cray we had in Houston when I worked for Chevron in the early 90s! Think about that, and then think about how little data you really need to deal with to do the job, vs what you're actually shuffling around.)