Let's say you have an idea for an app. You've written a brutally good prompt to the AI, down to the smallest detail in terms of:
What's the purpose of the app
Defining the functionality.
The user's role
-restrictions
-tech stack, etc.
But you didn't give him enough 𝗰𝗼𝗻𝘁𝗲𝘅𝘁.
If the AI does not have information in front of it:
-How the project is organized and what kind of architecture it uses.
-What rules and conventions are already in place?
-How the frontend, backend, and database are connected.
-What are the rules of business that need to be followed?
-What features have already been implemented and how they work.
-which libraries to install and what to leave out
-which rules to stick to
It can easily happen that it generates an AI slop for you that later has to spend a lot of time to fix it. It can generate code to some extent that works half-as we expected, but that would be chaotic for further maintenance, especially for production.
So, before an agent writes you the first line of code, make sure you have good context and a good prompt afterward.
Documents such as PRD.md, Tech.md, DB.md and files similar to them are a solution to this, as this extremely reduces the chance of bad output.
Linkedin: https://www.linkedin.com/in/lazar-a%C4%87imovi%C4%87-82a372301/
Top comments (0)