One Month of Vibe Coding with GitHub Copilot

Yingting Huang -

Today is the last day of October, and I just finished a one‑month challenge: using GitHub Copilot for vibe coding. Along the way, I also tried Codex, GitHub Copilot CLI, and Spec Driven Development.

In this one month, I built four apps:

  • Underwriting X
  • Claim X
  • Reading X
  • Document Generator

Here are my main lessons from this journey.

1. Fundamentals Still Matter

The basics never change: Algorithms + Data Structures = Programs.
Before asking AI to generate code, define your data model first. Once the model is clear, Copilot can create useful code around it.

2. Plan First, Guide the AI

You are still in control. Don’t let AI decide everything. Clear your mind, plan your steps, and then guide Copilot.

3. Ask Questions

If you don’t know what to do, keep asking. AI can act like a teacher and give you advice. Then you can decide the next move.

4. Be Clear with Instructions

Vague ideas lead to messy code. If your input is not clear, Copilot may generate random code that is hard to refactor. Define clear input and output to avoid chaos.

5. Document with AGENTS.md

Create an AGENTS.md file to describe roles and responsibilities of different parts of your system. This helps Copilot understand your project better.

6. Prototype First, Spec Later

When building a prototype, don’t waste time writing specs. Ideas change fast. Just work with Copilot in the loop. Later, you can ask Copilot to generate documents for you.

7. Use Diagrams for UI

Diagrams are powerful. They help Copilot understand your UI ideas much better than text alone.

8. Choose the Right Model

Different models are good for different tasks:

  • Claude Sonnet 4.5: great for generating UI and specific modules.
  • GPT‑5 Codex: better for global reasoning and algorithms.

9. Keep Good Git Habits

Always commit your code and use branches. Sometimes Copilot generates broken code. With Git history, you can restore or roll back easily.

Final Thoughts

Copilot is not here to replace you. It is here to work with you. If you plan well, define your data models, and give clear instructions, AI can speed up your coding and even teach you new things.