I still think we don’t have a proper language for AI. Computers are deterministic, while LLMs are probabilistic models. Traditional programming languages are designed to avoid ambiguity, but LLMs try to mimic natural language, which adds another layer of ambiguity.
What I’ve noticed, though, is that people using vibe coding focus more on the final artifact, almost like product managers and that’s actually a good thing, especially for prototyping.
PS
Regarding progress in AI languages, I’ve been following Domingos’ Tensor Logic development (
https://homes.cs.washington.edu/~pedrod/tls.pdf). It tries to unify symbolic AI with deep learning.