<Bold_Prompts> #9

How to build your prompts backwards

Reveka Backtrackster will help you reverse engineer prompts.

Welcome to the 9th issue of <Bold Prompts>: the weekly newsletter that sharpens your AI skills, one clever prompt at a time.

Every week, I send you an advanced prompt inspired by a real-world application. Think of these emails as mini-courses in prompt engineering.

Today we’ll talk about building your prompts backwards.

Reverse engineering is the art of starting from the finish line and working your way back to the first step. It’s a technique engineers use to break down and understand complex technology — be it code or hardware.

And guess what?

You can apply the same principle to your prompts.

Instead of writing a prompt from scratch, you can start with a high-quality output and navigate your way back to a prompt.

Reverse prompt engineering is a viable technique because of a specific side effect that comes from training LLMs.

You see, LLMs train on thousands of {question; answer} snippets to learn how to respond to instructions in the same way a human assistant would.

And it turns out LLMs didn’t only learn to write better answers; they also learned to formulate better questions. Reverse prompting is all about leveraging this “side-effect” ability.

The technique shines in two situations:

1. When you need inspiration to craft a detailed prompt — like a prompt meant to generate synthetic data or apply a long list of steps.

2. When your goal is to generate output with very specific formats — like a JSON object, a board game, a landing page, or a recipe.

Today’s prompt covers both cases with a specific methodology. You’ll be able to :

• Start from a given output (like a recipe) and generate a prompt that produces similar results (other recipes with a similar format but different content).

• Generate prompt templates and step-by-step instructions.

• Co-generate detailed prompts with your favorite LLM.

Alright, let’s dive right into it: