Designing Scalable Prompts

---
title: Scalable Prompt Design
tags:
- llm
- prompt engineering
search: true
toc: true
toc_label: My Table of Contents
toc_icon: cog
classes: wide
---
Separate stable concepts from changing details. Your prompts become maintainable, not messy.

Like a lot of developers, when I first started working with LLMs I wrote prompts in a very ad-hoc way. They worked… until they didn’t. Every new requirement meant I had to re-write or patch together instructions. It quickly turned into prompt soup — hard to maintain, inconsistent, and fragile.

That got me thinking: what if I applied the same design principles I use when writing software?

Specifically, I looked at the SOLID principles. These ideas have been around for decades in software engineering, but it turns out they map really well to prompt engineering too. While some principles are not relevent, I still managed to have a simple but powerful lesson:
separate the stable concepts from the changing implementation.

## From Messy Prompt to Structured Design

Here’s an early example from my practice.

I started with something like this: