Git, GitHub and AI Tools

Learn the tools that are changing how researchers code

Halbritter Lab / CeRKiD | February 17, 2026

Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

What You'll Learn Today

By the end of this session, you'll be able to:

  • Understand Git basics: repos, commits, branches, and PRs
  • Protect your research: .gitignore, private repos, and PII management
  • Explain why version control is essential when using AI tools
  • Navigate the AI tool landscape and choose what fits your workflow
  • Follow the complete Git workflow hands-on with KidneyQuest
Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

We've All Been Here

final_v3.docx
final_v3_REALLY_final.docx
final_v3_REALLY_final_reviewed.docx
final_v3_REALLY_final_reviewed_comments_addressed.docx

Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

Version Control Unlocks AI's Potential

  • And it's not just about avoiding filename chaos
  • It's the foundation for working effectively with AI
  • Version control lets you experiment, fail safely, and iterate fast
Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

Reproducibility Requires Version Control

Scenario: You need to reproduce results from 6 months ago.

  • Which version of the script did you use?
  • What were the exact parameters?
  • Which dataset version?

Without version control: Good luck.
With version control: git checkout and you're there.

Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

Part 1: Git Basics

Your foundation for version control

Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

A Repository Tracks Your Entire Project History

Think of it as a lab notebook for your code.

  • Every change is tracked automatically
  • You can see who changed what and when
  • You can jump back to any previous state

A repository (or "repo") is just a folder with a .git directory that stores the history.

Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

Every Commit Is a Safe Restore Point

  • A commit is a save point
  • It captures the exact state of all files at that moment
  • Each commit has a unique ID and a message describing what changed

Think of commits as save points in a game. You can always go back.

Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

Good Commit Messages Tell Your Project's Story

Bad commit message:
"fixed stuff"

Good commit message:
"fix: correct gene ID validation to handle ENSG format"

  • Commit messages tell the story of your project
  • They help you (and others) understand what changed and why
  • Write them for your future self
Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

Conventional Commits Make History Machine-Readable

Format: type(scope): description

Common types:

  • feat: - New feature
  • fix: - Bug fix
  • docs: - Documentation
  • refactor: - Code restructure
  • test: - Tests

Example: feat: add gene search filter

Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

Semantic Versioning Tells You What Changed

v1.2.3 means:

  • 1 (MAJOR) - Breaking changes
  • 2 (MINOR) - New features, backward compatible
  • 3 (PATCH) - Bug fixes

Why researchers care: Reproducibility

  • Cite specific versions in papers
  • Know when updates break things
Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

Private Repos Protect Unpublished Research

Public repositories:

  • Anyone can view, clone, and contribute
  • Good for: Open source, teaching, published tools

Private repositories:

  • Restricted access only
  • Good for: Unpublished research, patient-adjacent data, pre-publication code

Default for medical research: Start private, make public when ready

Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

.gitignore Prevents Accidents Before They Happen

What it is: A file telling Git which files to never track

What to ignore:

  • Large data (CSVs, FASTQs, imaging)
  • Environment files (.env, API keys)
  • Binary files (PDFs, compiled code)
  • OS artifacts (.DS_Store, Thumbs.db)

Why it matters: Prevents accidental commits of things that should never be in a repo

Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

Never Commit Patient Data or Credentials

The rules:

  • No patient data, even "anonymized"
  • No API keys, passwords, or tokens
  • No institutional credentials

What if you do? It's in the history forever. Rewriting history is painful.

Best practice: Always review diffs before committing

Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

If you accidentally commit a password to GitHub, can you just delete the file?

Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

Branches Let You Experiment Without Risk

  • A branch is like running a parallel experiment
  • The main branch stays safe and stable
  • You work on your branch, test your changes, then merge when ready

Example:

  • main branch: the stable version
  • add-visualization branch: your new feature
Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

Pull Requests Are Where Collaboration Happens

  • You've finished work on your branch
  • Now you want to merge it into main
  • You open a pull request (PR)

What happens:

  • Someone reviews your code
  • You discuss, make changes if needed
  • Once approved, you merge

This is how teams collaborate.

Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

Git Is Local, GitHub Is Online

Git:

  • The version control tool on your computer
  • Tracks changes, creates commits, manages branches
  • Works offline

GitHub:

  • A website for hosting Git repositories
  • Enables collaboration, pull requests, issue tracking
  • Where you share your code
Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

Part 2: Version Control in the AI Era

Why Git matters more when AI writes your code

Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

AI Is Fast - Version Control Keeps You Safe

  • AI tools can write hundreds of lines of code in seconds
  • You need to track what changed and why
  • You need to review AI-generated code carefully

Version control gives you:

  • A clear history of what the AI added
  • The ability to roll back if something breaks
  • Confidence to try bold experiments
Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

Experiment Freely, Fail Safely

  • AI makes suggestions. Some are great. Some are wrong.
  • With version control, you can try anything
  • If it doesn't work, git reset and you're back to safety

The combination: Git + AI = you can be bold

Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

AI Coding Tools Evolved Rapidly Since 2022

  • November 2022: ChatGPT launches, goes viral
  • 2022-2023: GitHub Copilot gains adoption
  • 2023: GPT-4, Claude, Gemini released
  • 2024-2025: Cursor, Claude Code, Windsurf emerge
  • 2026: AI writes significant portions of production code

The pace is accelerating. That's why we're here today.

Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

Part 3: The Tool Landscape

Navigate AI coding tools with confidence

Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

Three Categories of AI Coding Tools

Three categories:

  1. Chat-based: ChatGPT, Claude.ai
  2. IDE sidebar: GitHub Copilot, Cursor
  3. CLI agents: Claude Code, Codex CLI

Each has strengths. Choose based on your workflow.

Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

Choose the Tool That Fits Your Workflow

For today's hands-on session:

  • You can use any tool you prefer
  • Chat-based is fine
  • Sidebar is fine
  • CLI is fine

No single right answer. Use what feels comfortable.

Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

Chat Tools Are Best for Learning and Quick Questions

Examples: ChatGPT, Claude.ai, Gemini

Good for:

  • Explaining concepts
  • Writing small code snippets
  • Debugging errors
  • Planning architecture

Limitation: You copy/paste code manually.

Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

Sidebar Tools Integrate AI Into Your Editor

Examples: GitHub Copilot, Cursor, Cody

Good for:

  • Autocomplete while you type
  • Inline suggestions
  • Editing existing code
  • Staying in your editor

Benefit: Tight integration with your workflow.

Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

CLI Agents Automate Entire Tasks

Examples: Claude Code, Codex CLI, Aider

Good for:

  • Autonomous tasks ("build a login page")
  • Multi-file refactoring
  • Following complex instructions
  • Running terminal commands

Tradeoff: Less control, more automation.

Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

The Course Site Has Everything You Need

The AI Tools page on the course site has:

  • Detailed comparisons
  • Setup instructions for each tool
  • Pros and cons
  • Recommendations based on experience level

Link: https://halbritter-lab.github.io/AI-Teachathon/

Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

Part 4: What We'll Do Today

From clone to merge in five steps

Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

KidneyQuest Teaches the Full Git Workflow

A simple browser game where the CeRKiD zebra collects genes.

  • Built with HTML, CSS, and JavaScript
  • Runs in your browser, no server needed
  • Simple enough to complete in 2 hours
  • Real enough to teach you the workflow
Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

Five Steps from Clone to Merge

  1. Clone the repo: Get the code on your computer
  2. Create a branch: Work safely without affecting main
  3. Code with AI assistance: Add a feature or fix a bug
  4. Open a Pull Request: Propose your changes
  5. Review and merge: Collaborate with a partner

Each step is simple. This is easier than it looks.

Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

Steps 1-2: Get the Code and Create Your Branch

Clone the repository:
git clone https://github.com/berntpopp/KidneyQuest.git

Create your branch:
git checkout -b add-new-gene

Now you have the code and your own safe workspace. Main is untouched.

Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

Step 3: Let AI Write Code While You Steer

Example prompt to your AI tool:

"Add a scoreboard to KidneyQuest that displays the number of genes collected"

What the AI might generate:

  • HTML for the scoreboard element
  • JavaScript to track and update the score
  • CSS for positioning and styling

Your job: Review the code, test in the browser, iterate

You are the pilot. AI is the copilot.

Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

Step 4: Propose Your Changes with a Pull Request

  • Push your branch to GitHub
  • Open a PR from your branch to main
  • Write a description of what you changed

This is where collaboration happens.

Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

Step 5: Review, Discuss, and Merge

  • A partner reviews your PR
  • You discuss, make changes if needed
  • Once approved, merge into main

Congratulations. You've contributed to the project.

Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

What You Now Know

You've learned:

  • Git tracks every change safely - repos, commits, branches
  • Branches and PRs let you experiment and collaborate without risk
  • .gitignore and private repos protect sensitive research data
  • AI + version control = safe, fast iteration
  • The workflow: clone, branch, code with AI, PR, merge

Next up: Apply all of this hands-on with KidneyQuest

Git, GitHub and AI Tools
AI-Teachathon | Halbritter Lab

Let's Get Started!

Course site: https://halbritter-lab.github.io/AI-Teachathon/

Git, GitHub and AI Tools

Welcome everyone. We're going to spend the next 30 minutes talking about version control, AI tools, and why they work so well together. Then we'll jump into hands-on coding.

This primes the audience for what matters. Keep to 5 objectives max. Use action verbs. Revisit these at the end to close the loop.

ENGAGEMENT: Show of hands - who has more than 5 versions of a file right now? [Pause for laughter/recognition] This is the traditional approach to version control. You know this system. It works until it doesn't. The problem: which version has the correct analysis? Which one did you send to the collaborator?

That's the hook. Version control is how you unlock AI's potential. You need to track what the AI changed, roll back when it makes mistakes, and maintain a clear history of your project.

This is not hypothetical. Reviewers ask for this. Reproducibility is essential. Version control makes it possible.

From here on, we use real terms, not research analogies. A repo is a folder with history. That's it.

Commits are the building blocks of Git. You make changes, stage them, commit them with a message. Simple.

You'll thank yourself later. Trust me.

This builds on the previous slide. KidneyQuest uses this format. It enables automated changelogs and clear history. Show a real example from the repo's git log if possible.

Bumping from v1.x to v2.0 means "this may break your analysis." Link back to conventional commits: feat bumps MINOR, fix bumps PATCH.

Charite context: most research should be private until publication. GitHub gives you unlimited private repos for free.

.gitignore is your first line of defense. Use GitHub's templates for your language. Prevention is easier than cleanup.

For medical research, this is non-negotiable. Prevention is everything. A single accidental commit of patient data can be a serious compliance issue.

Give 10 seconds of silence. Then explain: No - it's in the history forever. Even if you delete the file in a new commit, the old commit still has it. This is why .gitignore and reviewing diffs matters. Prevention is everything. This reinforces the previous 3 slides.

This is the second and last research analogy. After this, we use real terms. Branches let you try things without breaking the main project.

PRs are where collaboration happens. Code review. Discussion. Then merge. This is the workflow you'll follow today.

ENGAGEMENT: Quick check - "If your laptop dies right now, is your Git history lost? What about your GitHub repos?" [Pause for 2-3 responses. Answer: Git is local (lost if not pushed), GitHub is remote (safe).] Brief, clear distinction. Git is the tool. GitHub is the platform. You need both.

AI is powerful but not perfect. Version control is your safety net.

This is the key insight. Version control unlocks AI's potential because you can experiment without fear.

ENGAGEMENT: Show of hands - who has used ChatGPT, Copilot, or any AI coding tool before? [Gauge the room's experience level. Adjust depth accordingly.] This is moving fast. Two years ago, GitHub Copilot was cutting edge. Now we have AI agents that write entire features. You need to keep up.

This is the overview. Next slides break down each category.

The goal is to learn the workflow, not master a specific tool. Pick what works for you. We'll briefly cover each category next.

Chat tools are great for learning and quick questions. Not ideal for writing large codebases because of the copy/paste friction.

Sidebar tools are the sweet spot for many developers. Suggestions appear as you type. You stay in your editor. This is what most people use daily.

CLI agents are powerful but require trust. They can modify multiple files, run commands, even commit to Git. Advanced users love them. Beginners should start with chat or sidebar.

We won't cover every detail now. The site has everything. Bookmark it.

This is the hands-on project. It's simple but it covers all the concepts: cloning, branching, coding with AI, opening a PR, reviewing, and merging.

ENGAGEMENT: Turn to your neighbor - which step do you think will be the hardest? [30 seconds. Then ask 1-2 pairs to share. Common answer: Step 3 or Step 5.] This is the workflow. Five steps. You'll repeat this pattern for the rest of your career. It's the foundation of modern software development.

Two quick commands, paired together. We'll walk through this together. No one gets left behind.

This is the most important step. Show a concrete example so they see what AI-assisted coding looks like. AI will make mistakes - that's normal. Review, test, iterate.

PRs are the core of team workflows. You'll do this dozens of times in the hands-on session. Practice makes perfect.

This is the full cycle. Clone, branch, code, PR, merge. You'll repeat this pattern for every feature.

Bridge between theory and practice. Reinforce key takeaways. Point back to the objectives from the beginning - they've now covered each one.

Time to jump into the hands-on session. You've got this.