1 Prerequisites & Philosophy
Before we install anything, let’s understand what we’re working with and why it matters.
1.1 What is Claude Code?
Claude Code is a command-line interface (CLI) tool made by Anthropic that lets you interact with Claude—a large language model—directly in your terminal. Unlike ChatGPT or the Claude web interface, Claude Code can:
- See your files: It can read your entire project, understand the structure, and reference specific code
- Make changes: It can edit files, create new ones, and delete what’s not needed
- Run commands: It can execute terminal commands—build your project, run tests, commit to git
- Work iteratively: You have a conversation, refine your request, and watch it work
Think of it as having a collaborator sitting next to you who can type very fast and knows a lot about many things—but still needs your guidance and approval.
1.2 Why Claude Code for Academics?
You might wonder: “I’m not a software developer. Why would I use a terminal-based coding tool?”
Here’s the thing: much of academic work is computational, even if we don’t think of it that way.
| Task | Traditional Approach | With Claude Code |
|---|---|---|
| Format a bibliography | Manually fix each entry | “Fix the formatting of all BibTeX entries to match Nature style” |
| Analyze survey data | Click through SPSS menus | “Load this CSV, run a correlation matrix, and plot the significant relationships” |
| Make a figure | Fiddle with matplotlib for hours | “Make this figure publication-ready: increase font size, use colorblind-safe palette” |
| Install a tool | Google error messages for hours | “Install samtools and fix whatever errors come up” |
The terminal isn’t a barrier—it’s a direct line to getting things done without clicking through menus.
1.3 What You’ll Need
1.3.1 Required
- A computer: Mac, Windows, or Linux all work
- Internet connection: Claude Code communicates with Anthropic’s servers
- Time: Set aside 1-2 hours for initial setup, then learn as you go
1.3.2 Accounts You’ll Create
- Anthropic account: To access Claude (you’ll get an API key)
- GitHub account: For version control and collaboration (optional but recommended)
1.3.3 Cost Considerations
Claude Code uses Anthropic’s API, which has usage-based pricing. As a student or academic:
If you’re attending a workshop, API access may be provided. Check with your instructor.
- API credits: Anthropic occasionally offers free credits for academic use
- Typical costs: Light usage might cost $5-20/month; heavy usage more
- Free alternatives: Some institutions provide access; ask your IT department
We’ll discuss this more in the setup section.
1.4 Philosophy: How to Learn with AI
Working with Claude Code is different from traditional learning. Here are principles that will help:
1.4.1 1. Start with Intent, Not Syntax
Don’t worry about how to do something. Start with what you want:
“I want to download all supplementary tables from this paper’s DOI”
Claude Code will figure out the how. Your job is to be clear about the what.
1.4.2 2. Iterate, Don’t Perfect
Your first prompt won’t be perfect. That’s fine. Watch what Claude does, then refine:
- “Analyze this data” → Too vague, results aren’t useful
- “Run a PCA on columns 3-10 and plot the first two components” → Better
- “…and color the points by the ‘condition’ column, add a legend” → Getting there
1.4.3 3. Trust but Verify
Claude Code is powerful but not infallible. Always:
- Review changes before accepting them
- Understand commands before running them (especially anything with
rm,sudo, or that touches important files) - Check outputs make sense for your domain
1.4.4 4. Learn by Watching
One of the best ways to learn is to watch what Claude Code does:
- It will show you the commands it runs
- It will explain its reasoning
- You’ll pick up patterns and eventually do some things yourself
1.4.5 5. It’s Okay to Not Know
You don’t need to understand every line of code Claude writes. Focus on:
- Does the output match what I wanted?
- Can I explain at a high level what happened?
- Would I know how to ask for changes?
1.5 What We’ll Build Together
By the end of this course, you’ll have completed at least one of these projects:
Meta-analysis of scientific tools: Programmatically searching literature, extracting data, and visualizing which tools are most used in practice
Automated citation insertion: Taking a draft paper and systematically finding and inserting appropriate references from PubMed
Protein structure visualization: Using PyMol to create publication-quality figures of binding sites—without prior PyMol knowledge
Each project teaches transferable skills. The specific topic matters less than the approach: breaking down problems, working iteratively with AI, and producing real outputs.
1.6 Ready?
Let’s set up your environment. Continue to Part 1: Installing Visual Studio Code.