Part 3: GitFlow + AI Development Tool-Based Collaboration Practice

Course Overview

This course covers Git collaboration workflows needed when transitioning from individual to team projects.

We start not by listing GitFlow rules, but by understanding why such a structure is necessary.

In particular, we emphasize that in an era where AI coding tools have become ubiquitous, Git goes beyond simple version control to serve as a “safety net for AI-generated code.”

Learning Objectives

• Fundamentally understand why branch strategies are necessary
• Grasp each GitFlow branch’s role and flow
• Learn how Pull Requests and code reviews actually work
• Learn how to leverage AI tools for commit messages and review assistance
• Internalize the development cycle connecting AI + testing + GitFlow

Course Structure

Part 1: Why Branch Strategies Are Necessary

• Problems created by “pushing directly to main”
• Differences between solo development and team development
• Branches = work isolation + merge control
• Real conflict cases from actual team projects

Part 2: Understanding GitFlow as a Mental Model

• Roles of main, develop, feature, release, and hotfix
• What each branch “protects”
• GitFlow vs GitHub Flow vs Trunk-based: When to use what
• Simplified GitFlow suitable for labs and undergraduate teams

Part 3: The AI Era Development Cycle — AI + Testing + Git

In an era where AI coding tools generate code, Git is not just a version control tool. It serves as a “checkpoint” that allows you to safely revert when AI-generated code doesn’t work as expected.

flowchart TD
    A[Create feature branch<br/><br/>git checkout -b feature/new-function]
    B[Request code from AI<br/><br/>'Write a function to validate user input']
    A --> B
    
    C[Run tests<br/><br/>pytest tests/]
    B --> C

    C --> D{Test results}
    
    E[commit<br/><br/>Proceed to next task]
    D -->|Pass| E

    F[git checkout -- .<br/><br/>Re-request from AI with error message]
    D -->|Fail| F
    F --> B

Why This Cycle Matters

AI-generated code often “looks plausible but is wrong.” Accumulating AI code without testing makes it difficult to trace where problems started. Combining Git’s version control with testing enables:

• Failure isolation: Revert only the failing changes when tests fail
• Safe experimentation: Try AI code freely on branches, discard if it doesn’t work
• Incremental integration: Merge only validated code in small units to develop

Practice: Experience the AI + Testing + Git Cycle

• Request function implementation from AI
• Validate with pre-written tests
• On failure, revert with git checkout and retry
• On success, record with meaningful commit message

Part 4: Pull Requests and Code Review

• Why PRs are more than just merge requests
• Components of a good PR: title, description, scope of changes
• What to look for and what not to look for in code review
• Communication tips for both reviewers and authors
• Special considerations when reviewing AI-generated code

Part 5: AI-Based Collaboration Tools

• Auto-generating commit messages with AI
• Using AI to draft PR descriptions
• Introduction to AI code review assistance tools
• Limitations of AI tools and the role of humans

Part 6: Real Team Project Simulation

• Conduct a virtual project in teams of 3-4
• Create feature branch → Write code with AI → Test → PR → Review → Merge
• Practice resolving conflicts
• Retrospective: What went well and what was difficult

Course Format

• Online/Offline: Zoom or in-person sessions
• Hands-on focused: Team collaboration simulation in actual GitHub repositories

Target Audience

• Undergraduates starting their first team project
• Teams preparing graduation projects or capstone projects
• New lab members working with shared codebases
• Developers who have only used Git solo without collaboration experience

Prerequisites

• Basic Git commands (add, commit, push, pull)
• GitHub account and basic usage experience
• Ability to write simple code in Python or another language
• Basic pytest usage (TDD course completion recommended)

Key Practice Examples

• Creating feature branches and working on them
• AI-generated code → test → commit/rollback cycle experience
• Writing PRs and exchanging reviews
• Resolving merge conflicts
• Generating commit messages with AI tools

AI Era Development Cycle Summary

GitFlow Branch Roles Summary

Good Commit Message Structure

<type>: <subject>

<body>

<footer>

• type: feat, fix, docs, style, refactor, test, chore
• subject: Under 50 characters, written in imperative mood
• body: Explain why this change is needed
• footer: Link issue numbers (e.g., Closes #123)

Next Steps After This Course

After completing this course, you can continue with the DevOps / MLOps Minimum Set course, where you’ll learn how to scale team-level collaboration to organization-level stable operations.

Contact

For inquiries about course schedules and pricing, please reach out via email.