Today, I made a Python script that takes your prompt and generates a playlist on Spotify based on that prompt.
Colt Steele, the instructor, explained how he implemented the playlist generator. I did not code along. Instead, I listened, made notes and tried to understand. That was yesterday. Today, I did a closed book re-implementation and it worked.
Lessons learned:
- leverage the LLM's ability to respond with structured output
- prompt engineering is not dead
- speaking of prompt engineering: when you get a good response from the LLM, copy it into the prompt as an example to follow
- most of the coding is the glue between the LLM's response and the service's API
- UX is not dead: I showed my playlist generator from the terminal to my girlfriend. she shuddered and said: "no way am I ever going to use that" 🤷♂️