[pinkcan]

1. the script is generated by the llm

2. the user runs the script that does the scraping

these are temporally separate actions

[dopidopHN]

Fine, but it’s subject to html selectors brittleness no? Oh, you subject the raw html when you need it maybe?

[dagurp]

Here's how I do it.

1. Tell chatGPT to create a python script that scrapes example.com and generate an rss file.

2. Paste a snippet of the html and tell it to modify the script to use that.

3. I do some minor tweaks myself to fix the date format.