Engineering · Week 2
From "Build a CRUD API" to an Engineered Prompt
How I moved from giving an AI coding assistant a simple request to treating it as an engineering collaborator.
During Week 2 of my backend engineering work, I used a CRUD API assignment as an experiment in AI-assisted engineering.
The goal was not simply to get an AI system to generate code. I wanted to understand what happens when the AI is given enough engineering context to produce a more deliberate implementation.
The starting point
A simple request
The request communicates a broad goal, but leaves many engineering decisions unspecified. The implementation could vary considerably depending on the assumptions made by the AI.
The prompt ladder
Adding engineering context
I progressively refined the request by adding context, technical constraints, structure, output expectations, and quality requirements.
01 · Clear goal
Define exactly what the API should accomplish.
02 · Technical context
Specify the technologies and constraints that should guide the implementation.
03 · Define the structure
Give the AI clearer expectations about how the solution should be organised.
04 · Specify the output
Make the expected result explicit instead of leaving important implementation decisions open.
05 · Treat the prompt as a specification
Move beyond simply asking for code and provide enough context for the AI to reason about the engineering task.
What changed
The prompt became part of the engineering process
The experiment changed how I think about AI-assisted development. The quality of an AI-generated implementation depends heavily on the quality of the engineering context provided to the model.
Effective AI-assisted engineering is less about asking AI to write code and more about giving it a clear, reusable engineering specification.
The lesson
AI can accelerate implementation, but it does not remove responsibility
Generated code still needs to be inspected, tested, understood, and evaluated by the engineer. The role of the engineer is not removed by using AI; the engineer becomes responsible for directing and validating the implementation.