What did building an "LLM chat UI" teach me?

There's a difference between what I read about LLMs and generative AI in general, and what I experience when building things on top of the LLM API with the help of an LLM itself.

Also, over the past few weeks, I've realized that even if the development of language models stopped today, programming has already changed for me.

I wrote a few words about this, so if you work with LLMs in any way, this perspective might be interesting for you.