Language is really powerful, I think it's a huge part of our intelligence.
The interesting part of the article to me is the focus on fluency. I have not seen anything that LLMs do well that isn't related to powerful utilization of fluency.
I don't think changeable code is the number one priority. The goal is to solve a problem and code that solves a problem without needing to change is sufficient.
Code that doesn't need to change is a really good sign that you've got something good.
In the article the author mentions wanting to benchmark a GPU and using ChatGPT to write CUDA. Benchmarks are easy to mess up and to interpret incorrectly without understanding. I see this as an example where a subtly-wrong idea could cause cascading problems.