Why prompts are more akin to source code than simple instructions

Prompts used in large language models (LLMs) are more similar to source code than simple instructions, requiring integration into applications like Vertesia, which streamlines this process. The article explores how prompts, when used to execute tasks or act as assistants in applications, require elements such as variables and conditions, similar to programming. This complexity can lead to maintenance challenges akin to "spaghetti code" if not managed properly. Vertesia addresses these challenges by treating prompts as integral components of an application's architecture, much like source code, and provides tools to simplify the creation of LLM-powered applications. This approach involves composition, rigorous testing, structured data handling, and abstraction of low-level APIs, allowing applications to harness the power of LLMs to process data and present refined outputs, akin to how SQL is used in databases without direct user input.