> My bet? The language with the best feedback loop to LLMs wins. TypeScript has a huge advantage here. Static types give LLMs clear signals about correctness.
Tencent designed their own methodology of testing LLMs across languages. Elixir topped the lot on AutoCodeBench in terms of problems solved correctly:
Correct me if I'm wrong, but isn't Phoenix suppose to be for backend driven applications, similar to Ruby on Rails? Wouldn't trying to blend that with React always be a bit of a pain in the ass regardless of language?
Enjoyed this balanced perspective on Elixir.
In regards to this:
> My bet? The language with the best feedback loop to LLMs wins. TypeScript has a huge advantage here. Static types give LLMs clear signals about correctness.
Tencent designed their own methodology of testing LLMs across languages. Elixir topped the lot on AutoCodeBench in terms of problems solved correctly:
https://github.com/Tencent-Hunyuan/AutoCodeBenchmark?tab=rea...
An alternate bet might be that context efficiency wins. The more expressive, comprehensible, and idiomatic languages have the advantage.
Correct me if I'm wrong, but isn't Phoenix suppose to be for backend driven applications, similar to Ruby on Rails? Wouldn't trying to blend that with React always be a bit of a pain in the ass regardless of language?