Hacker News new | ask | show | jobs
Web Scraping with GPT-4 (jamesturk.net)
5 points by tomberin 1189 days ago
1 comments

This is fantastic. You're essentially using an LLM as a function call, passing in the HTML alone! I'll still probably use something like Puppeteer, or Bright Data (if needing low-code/no-code templates) for dynamic content, but for everything else this is a revelation, as a concept. The token limitation is a bother (and 32k context GPT-4 is going to be expensive) but manageable. Will definitely be exploring this option myself. Thanks!