Jorge Arango gives a real-world approach to using large language models (LLMs) like GPT-4 to re-categorize and re-tag almost 1,200 blog posts on his website. Faced with the challenge of enhancing the discoverability, AI is used to automate the identification and tagging of content for improved organization and search.

Key points:

• LLMs can automate time-consuming content re-categorization tasks by identifying relevant tags/categories based on the text

• Preparing a cleaned up, widely understood taxonomy is important for effective LLM tagging

• Generating a preview CSV allows reviewing LLM suggestions before applying changes

• Shell scripts can apply LLM outputs at scale by modifying local Markdown files

• LLMs may introduce new invalid terms, so human review and clean-up is still required