Getting Started¶
A 15-minute exercise you can try right now. No prior experience assumed.
Choose a platform¶
You need a free account on one of these:
| Platform | URL | Notes |
|---|---|---|
| Claude | claude.ai | Recommended. Strong for analysis and writing. |
| ChatGPT | chatgpt.com | Widely used. Good all-round. |
| Gemini | gemini.google.com | Google's offering. Good for web-connected tasks. |
Any of these will work for this exercise. If you already have an account somewhere, use that.
Don't Panic
You do not need a paid subscription for this exercise. The free tier on any platform is sufficient. You can always upgrade later if you find the tools useful.
The exercise (15 minutes)¶
Step 1: Upload a text (2 minutes)¶
Find a short text you know well — a passage from a primary source you teach, a paragraph from something you've written, or an extract from a key secondary source in your field. Copy and paste it into the chat window, or upload a PDF.
Step 2: Ask about it (3 minutes)¶
Try one of these prompts (adjust to your material):
- "Summarise the main argument of this text in three sentences, written for an undergraduate audience."
- "What are the three strongest objections a critical reader might raise against the argument in this text?"
- "Identify the rhetorical strategies used in this passage and explain how they serve the author's purpose."
Step 3: Evaluate (5 minutes)¶
Read the response critically. Ask yourself:
- Is it accurate? Does it correctly represent what the text says?
- Is it useful? Does it tell you anything you didn't already know, or frame something helpfully?
- Where does it go wrong? Most responses will contain at least one error, oversimplification, or misjudgement. Finding it is the exercise.
Step 4: Push back (5 minutes)¶
Now respond to the AI. Try:
- "That's not quite right about [specific point]. The text actually argues that..."
- "You've missed the most important aspect, which is..."
- "Good, but make the tone more formal and add specific textual references."
Notice how the response changes. This iterative refinement is how these tools are actually used — not as one-shot oracles, but as conversational partners that improve with specific feedback.
What you've just learned¶
Essential
- These tools respond to specificity. Vague prompts get vague answers. Detailed prompts get detailed answers.
- Your expertise matters. You spotted errors the tool didn't flag. A non-specialist would have missed them.
- Iteration is normal. First responses are starting points, not finished products.
- Verification is essential. The output sounded confident regardless of whether it was correct.
Basic prompting principles¶
For any task, specify:
- Audience — who is this for?
- Format — prose, bullet points, table, code?
- Length — how much?
- Scope — what to include, what to leave out
- Purpose — why do you need this?
For a full treatment of prompting strategy, see Prompting Principles in Essentials.
What to try next¶
- Use Cases — practical tasks organised by time investment
- Mindset — how to think about these tools
- Ancient Languages — if you work with translation and textual criticism
- How to Choose a Platform — if you're ready to pick a tool, or go straight to Claude or ChatGPT
- AI Basecamp — structured exercises for deeper learning
Leif's Notes
If you tried the exercise and thought "that was underwhelming" — that's a legitimate reaction, and it might mean these tools aren't useful for your particular work. But I'd suggest trying one more thing before deciding: upload a longer document (10+ pages) and ask the tool to identify patterns, inconsistencies, or connections across the whole text. That's where the scale advantage starts to show. The 15-minute exercise above is deliberately modest — it's designed to get you oriented, not to showcase the most impressive capabilities.