There have been times in the past where goofing around on my own projects or experiments, I've cranked out many hundreds of lines of code in a day. It's possible that I've exceeded 1000 lines in one day. It usually computed the task at hand at least once, but it wasn't commented, documented, code reviewed, fitted with unit tests, run through any kind of QA process, etc.
On the other hand, I would estimate that when working for a salary and writing code that ran on mission-critical systems, over the course of a project I probably averaged about 10 lines per day. Time spent actually writing the code was dwarfed by time spent negotiating architecture tradeoffs in meetings before even starting, testing the code, documenting it, writing test procedures, debugging issues generated by the test engineers, going through the revision and deployment workflow, reviewing other peoples' code, etc.
Which kind of code is being cranked out from a couple of chatbot prompts? My bet is that it is mostly like the first case I described, and it also probably works at least once.