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.