A very basic decision tree/random forest model. I haven't updated it since the first version I did about a year or two ago, because I didn't have users myself and no one provided any feedback. Thank you for the feedback! I've made a todo item to refresh/explain the model.

Looks like it uses github stars for that.

That would exclude many users for languages that are in legacy systems, which are rarely tracked in github.

It could be alright for a rough/relative estimate...

Or jobs?

I don't know, but looking at ADA, 20 jobs seems very low. I was under the impression that ADA is still a required language for a lot of military hardware.

My guess would be that he's scraping jobs sites for listings numbers, and I suspect something like ADA development is using different channels to find devs.

Yes, a year or two ago I grabbed job listings from sites like Indeed and LinkedIn and trained a very crude basket of models. No one ever seemed to use it so I hadn't touched it since then. Time to refresh that, thanks for the feedback! Interesting points about ADA and I'll try and bring in some of those back channels.

Prolog has 402 claimed jobs: https://codelani.com/languages/prolog.html but looking at any Prolog implementation lists 0. I cannot imagine that there are this many (available?) Prolog jobs in the world, nor that no advertised job would specify the implementation the company uses. I think the numbers are generated using some sort of garbage in, garbage out system.

See answer above. A very crude model. Will ship a new one with an explanation shortly.

Poorly, it seems.