8 Comments

This is an excellent and unique guide, thanks man! ✌🏼

Expand full comment

Thank you!

Expand full comment

Great article. I haven't encountered this approach elsewhere. I'll get my hands dirty and test it on my prompts now ;)

Good to see how rapidly your knowledge about LLM's evolved during last couple of months. Keep up the good work!

Expand full comment

Fantastic and full of great insights article! Loved every bit of it! Thanks!

Expand full comment

Structuring the prompt with html tags is a great idea. Thanks

Expand full comment

Fantastic, thanks for sharing.

Examples recommendations align well with the "Demonstrate" stage in the Stanford's DSP (Demonstrate-Search-Predict) framework.

Expand full comment

Thank you. Didn't know that one; I'll check it out.

Expand full comment

> In essence, the "Demonstrate" stage is about automatically generating comprehensive, pipeline-aware examples that show not just what you want to achieve, but also how to achieve it through a series of intermediate steps. This approach enables the framework to handle complex, knowledge-intensive NLP tasks more effectively than simpler "retrieve-then-read" pipelines.

Expand full comment