Prompt Length & API Limits

Huge Thanks for Sharing the Prompts!

First of all, thank you so much for sharing all the prompt files — it's super insightful and inspiring to see how cleverly structured prompting alone can result in such powerful tools like v0, Lovable, Cursor, etc.

I'm exploring the idea of building something similar, but scoped specifically to Salesforce Apex boilerplate generation, and I had a few questions on how to practically make use of these large prompt files.

Questions

1. Prompt Length & API Limits

Many of the prompt files go over 1000+ lines. If I wanted to build a similar product:

• How would I manage sending such long prompts via the API?

• Wouldn’t the combination of prompt + user code exceed GPT-4’s token limits?

2. Managing Large Prompt Files

What’s your recommended approach here?

• Should I split prompts into smaller, modular chunks and load them dynamically?

• Is vectorizing/indexing the prompt content (for retrieval) the ideal method?

• Any tools/libraries you used or recommend to manage this complexity?

3. Adapting Prompts to a Narrower Domain

If I’m building for a specific use case (e.g., only Salesforce Apex code):

• Would you suggest starting from scratch with fresh, focused prompting?

• Or should I trim/adapt your existing prompt structures to suit the domain?

4. Productizing & Open Sourcing

Any advice on turning such prompt-powered systems into reliable open-source tools like Cursor or Lovable?

• What was key in evolving from “just a prompt” to a polished product?

Would love to hear your thoughts on this — especially since your work already sets a high bar for quality in this space. Thanks again for being so open with your process and sharing it with the community!