Then when we want to search for things related to the project we can use its UUID instead of the name. In most cases I'd expect that to return relevant results with only a small number of unrelated hits.
Searching on a random UUID is interesting. I generated a version 4 UUID, cc8ece66-ded9-45f0-ad45-579bb97251bf, and searched for it in Google. It gets one hit, to a page on the "Death Road to Canada" wika at fandom.com [1]. Google specifically says that the page is missing cc8ece66-ded9-45f0-ad45-579bb97251bf, so I have no idea why it offers it.
Duck Duck Go gives a bunch of hits related to cryptocurrency and bittorrent stuff, many in Russian, plus some other Russian sites, and a few other odd things.
Bing says that there are no results.
For a couple other random UUIDs (6bf402ec-4428-418c-9bfc-f54257fa8b2f and 93c7e4a2-bdc2-4cd0-8afd-2351f7b9dd4d), results were similar. For those Bing and Google both had no results, and Duck Duck Go had results similar to those for cc8ece66-ded9-45f0-ad45-579bb97251bf.
My earlier comment was about webpages, not projects, though I think having a project-specific UUID would have major advantages as you point out.
> Google specifically says that the page is missing cc8ece66-ded9-45f0-ad45-579bb97251bf, so I have no idea why it offers it.
Looks like Google is matching only "ded9". I recall that Google treats dashes as word separators and underscores not as word separators. This might make a UUID not a good choice because you only want to display exact matches.
https://searchengineland.com/google-bing-handle-underscores-...
As a fluid dynamicist, I might expect TensorFlow to have something to do with fluid dynamics given that I use tensors and flows, but nope...