For all the people downvoting - despite what the clickbaity title might suggest this is not a slop post by some AI evangelist.
This is in fact an article by Elixir creator José Valim, where he comments on a recent study which showed that models had on average much better performance using Elixir to solve problems out of the 20 languages tested.
He makes some solid arguments to explain why that could be the case and why Elixir is particularly easy to work with for LLMs as well as humans.
I think I saw an article (not based on this study, and I'm not sure based on anything) that claimed a different language to be the same ‘best for AI’
I expect we may see more, and each time we may come up with some explanation of why, of course. Still maybe I should learn Elixir, it sounds decent
Oh it's a great language, I honestly haven't met anyone who started writing Elixir and didn't end up loving it. If you have the time I definitely recommend it.
@revolutionaryvole @sukhmel Yes, loving it. But can't use it in the job 😭
It's still a rare opportunity unfortunately, I count myself as extremely lucky to have used it in a previous job!
What did you use it for?
Our entire backend was Elixir-based.
There were a lot more components involved but I don't want to go into too much detail. It was a real joy to work with though, and I'll definitely choose Elixir for any future projects unless there are specific reasons requiring something else.
Thanks for the insight. I've started learning Elixir some months ago and while it is a little difficult to get into for me I can see that once you know the basics it can be very powerful. However I'm not a programmer so I'll probably only use it for hobby projects.
even if this isn't a slop post and the author is credible, timing my friend. LLMs being used for good things is by far the exception to the rule