The code that the prompt generators generate was always around - they are called 'boilerplate code'. Anyone who knew how to do some technical searches could access those from common repositories or even Stackoverflow. But this was something difficult to do for non-technical people, or even the people with some insight into programming, so they are treating the prompt generators as a 'major revolution'. Yeah, they are, if you are a non-technical person who just wants to make a small modification to your website to show your last X posts or last X backers on the front page. But in the actual programming scene, nothing has changed - the code that the prompt generators generate keeps existing as boilerplate code available from many repos and sites, and they may or may not be used in a larger code block depending on the circumstances. A lot of programmers prefer to just use existing boilerplate code because such code is publicly vetted and tested whereas you have to sit and read the prompt generator's code to make sure that it wont have some unintended effect - which is a higher cognitive cost than actually writing that code, leaving aside just using boilerplate code.

And the biggest elephant in the room is still unaddressed: Copyright. Its all fun and games until a copyright holder notices that their code has been used by LLMs to generate code which a lot of people use and then takes all of them to court. Having a lot of Open Source licensed code in the LLM data does not fix that - It takes just a bit of non-Open Source code to have been in the LLM data set to justify legal action. And you cant drop the ball in OpenAI etc's court either: The copyright holder would sue both of you and he would tell you to settle your beef with OpenAI separately.