Smaller prompts, better apps

Building prompt-based applications can be a complex task. There are many dynamic variables, lots of logic, and many product decisions and assumptions that we must make. In many cases, breaking down prompts into smaller consumable pieces can help teams build products and applications faster with higher quality.

Prompt Size

A good prompt is one that works. However, as we increase prompt complexity, the LLM will start to lose the understanding of text. In many cases, especially for larger prompt-based applications, breaking large prompts into smaller ones will significantly impact overall output quality.


Prompts should be written like functions. When we write a Python or JavaScript function, it is best practice to write them into small, logical components - instead of one giant one. This concept should also apply to prompts - small, logical components.

Prompt Management

Managing multiple prompts will always be more challenging than managing a single one. However, the alternative could lead to an inferior product and user experience. Many tools on the market exist today to assist with this, including PromptDesk. If an application requires a lot of logic, context, understanding, and reasoning, we should create many different prompts for those components - the same way we would write many source code functions.


Breaking down prompts into smaller, consumable components can positively impact cross-team collaboration and prompt building. Subject matter expertise is often required to provide maximum value - especially in healthcare, construction, legal, technology, marketing/sales, etc. Breaking down prompts and allowing non-engineers to experiment, build, and optimize prompts can help accelerate an organization's overall AI mission.

Justin Macorin | LinkedIn
Lead Machine Learning Engineer, Seismic

Justin is a Lead Machine Learning Engineer with Seismic, the leading global sales enablement platform. His mission is to help accelerate the adoption of AI and help engineers accelerate their development of prompt-based applications.