This is an absolutely horrible idea. I’m not questioning the technology choice. But as someone interested in their career, it makes no sense to focus on a language or technology that is not popular. It’s both bad from the recruiting side trying to get developers who are smart enough to care about their n+1 job and the developer side.

There are probably less code samples and let’s be honest this is 2025, how well do LLMs generate code for obscure languages where the training data is more sparse?

Maybe for you! That's your call. I'm also interested in my career.

I've had 3 Elixir jobs and 2 Rust jobs in the last 10 years. All were on real products, not vaporware. I learned a ton, worked with great people, and made real friends doing it.

Luck? Skill? Who knows. It's not impossible to work with the technology of your choice on problems you find interesting if you're a little intentional.

Nothing ever gets better if everybody just does what's already popular.

I’ve made a very lucrative career moving from .NET to BEAM. I don’t even work with it currently but the fact I’ve shipped it for some pretty niche systems shows versatility and consistently goes in my favour when getting hired.

I work at a company doing full-stack Elixir with most of our devs all heavily using AI as they please to augment their workflow, and our CTO was genuinely concerned that our main competitor, a Python shop, had a leg-up on us for this exact reason.

He spent time running benchmarks for 0-1 apps and all kinds of other metrics and found basically no appreciable difference in the speed or accuracy of AI at generating Elixir vs. Python. Maybe some difference, but honestly it just doesn't exist enough to matter.

About LLMs: I did last year's advent or code with Elixir and when I forgot to turn off copilot it had no trouble writing whole implementations of functions, even if I had a very idiosyncratic style.

Most code is boilerplate and that's where LLMs shine, I don't think this specific issue is very important.

> this is 2025, how well do LLMs generate code for obscure languages where the training data is more sparse?

You'd be surprised: https://github.com/Tencent-Hunyuan/AutoCodeBenchmark/blob/b1...

> But as someone interested in their career, it makes no sense to focus on a language or technology that is not popular

A: why in gods name B: Every language, every framework and every tech stack is 1 month to 5 years away from being legacy crap. Unless you're learning something like KOBOL it's better to be able to use a variety of languages and show that you can adapt.

Reminds me of the time I was on a team doing stuff in Erlang for no reason

>how well do LLMs generate code for obscure languages where the training data is more sparse

LOL. Speaking about absolutely horrible ideas ...