The Human Question

How to Build Apps with Claude

A practical guide for professors, students, and academic staff. No coding required. Just clear thinking, structured prompts, and the willingness to iterate.

Prompting is the new programming

For decades, building software meant writing code. Now, you can build a working app by describing what you want in plain English. The skill that used to require years of training has been replaced, in part, by something most professors already do well: writing clear instructions.

"You don't need to be a programmer to build useful apps anymore. You need to know how to ask for one."

Claude is a thinking partner. You describe a problem. It generates the code. You test it, refine it, and ship it. The key is structure. Prompts that get good apps share four parts.

R
Role

Tell Claude who it should be. "An expert front-end developer..."

T
Task

State what to build. "Build a single-file HTML grade calculator..."

C
Context

Describe the world. "For a college professor, used on a phone..."

O
Output

Define what to deliver. "One self-contained .html file. No frameworks."

What you can actually build

Three categories cover almost every academic use case. Click any card for ideas.

A working demo

This grade calculator is a live, working app. Add a few assignments below and watch it compute. The whole thing was generated by Claude from a single prompt — which you can see at the bottom.

Grade Calculator

Assignment Weight (%) Score (%)
Total weight: 0%. Aim for 100% before final submission.
The single prompt that built this
"Act as an expert front-end developer. Build a single-file HTML grade calculator for a college professor. Inputs: assignment name, weight percentage, score percentage. Outputs: weighted average and a letter grade. Add and remove rows. Mobile-friendly. THQ royal blue (#1E40AF). One self-contained .html file. No frameworks, no build step."

How to build your first app

Five steps from idea to shareable URL. None of them require code experience.

1

Start with a real problem

Don't build for the sake of building. Pick something annoying you do every week — calculating grades, parsing student emails, generating quiz questions. The smaller and more boring, the better the first project.

2

Describe it in plain English

Open a notepad. In two sentences, say what the app does, who uses it, and what it produces. If you cannot describe it that briefly, the app is too big. Cut it down.

3

Use the four-part prompt

Role, Task, Context, Output. Drop your two-sentence description into the Task slot. Add a Role ("Act as a front-end developer"), a Context (who, where, on what device), and an Output (single HTML file, mobile-friendly, no external dependencies).

4

Test it, then iterate

Save the HTML Claude gives you to a file. Double-click it to open in a browser. Try it. When something is off, tell Claude what's wrong in plain English. "The grade column is too narrow on phones." It will fix it. The first version is rarely the last.

5

Share it

Drop the file into a free GitHub Pages repo, or paste it into your Squarespace Code Block, or just email it to a colleague. Done. You shipped a real, working tool. Now go build the next one.

Starter prompt library

Copy any of these prompts into Claude. Adjust the bracketed parts. Each one builds a working app in under five minutes.

Common mistakes (avoid these)

Being too vague

"Make me a tool" gets you a generic page. Specify the user, the inputs, the outputs.

Not defining the output

Say "single self-contained HTML file" or you might get React with a build step you can't run.

Expecting perfection on attempt one

Iterate. The second prompt is always better than the first because you saw what the first produced.

Skipping the test step

Open the file in a browser. Try it on your phone. Bugs are easy to fix when you see them.

Hard-coding sensitive data

Never put real student names, emails, or grades into a public-facing app prompt.

Building in isolation

Show your draft to one student or colleague before you ship. Five minutes of feedback saves an hour of rework.

You can build your own tools now

This is bigger than a productivity hack. The cost of building a small, useful app dropped to zero. The only thing left is the willingness to try, the patience to iterate, and the clarity to describe what you actually want.

Copied to clipboard