Your First Thing That Runs
You write the prompt, Claude writes the code, you run it — and something actually works.
This is the lesson where you stop reading about building things and actually build one.
By the end of this lesson, you will have a script that runs on your machine, produces output you can read, and does something genuinely useful. That is the goal. Not a screenshot. Not a simulation. Something that works.
Pick Your Thing
Before you open Claude, pick one of these. They are all tiny, universally useful, and require no setup beyond a terminal or a browser console.
Option A: Tip calculator. You give it a bill amount and a tip percentage. It gives you the tip and the total. Useful every time you eat out.
Option B: Email summarizer prompt generator. You give it the subject and a few lines of an email. It generates a ready-to-paste Claude prompt that summarizes the email for you. Useful every time your inbox gets heavy.
Option C: File renaming guide. You give it a list of file names and a pattern (like "add today's date to the front"). It writes out the renamed versions. Useful for anyone who drowns in screenshots named image_001.png.
Pick the one that interests you most. They all work the same way.
Write the Prompt
Open claude.ai. Start a new conversation. Write one paragraph describing what you want.
Here is an example for the tip calculator:
"Write a short JavaScript script that takes a bill amount and a tip percentage and calculates the tip amount and the total. Print the results clearly. Use a simple example — a $45 bill at 20% — to show the output."
Notice what that prompt does: it names the specific thing (tip calculator), the inputs (bill amount, tip percentage), the outputs (tip amount and total), the language (JavaScript — runs in any browser), and a test case so Claude knows what "working" looks like.
You do not have to use those exact words. What matters is the specificity. The more concrete your description, the more useful the output.
What Happens Next
Claude will produce a script. Read it quickly — not to understand every line, but to confirm it addresses what you asked for. Then copy it.
To run it in a browser console: Open any webpage in Chrome, Edge, or Firefox. Right-click → Inspect → Console tab. Paste the script. Press Enter. The output appears. If the browser warns you about pasting, type allow pasting and press Enter first — it is a one-time safety check — then paste the script. (Safari hides Inspect by default; turn it on under Settings → Advanced → "Show features for web developers," or just use one of the browsers above.)
To run it as a file: Paste the script into a plain text file, save it as script.js, open a terminal, navigate to where you saved it, and type node script.js. (Node.js is a free download from nodejs.org — it is not pre-installed on most computers. If typing node gives you a command not found error, either install it from nodejs.org or ask Claude for a version that runs directly in the browser console instead.)
Either way, within a minute of getting Claude's output, you will see your result on screen.
What You Are Looking At
When you see the output, look at the code Claude gave you. Every script it produces has the same three parts, no matter what the task is.
The input is the data you start with. For the tip calculator, that is the bill amount and the tip percentage. These are the numbers you would change to make the script useful for different situations.
The logic is the transformation — the part that actually does the work. One or two lines of math, or a function call, or a text operation. This is what Claude wrote that you did not have to.
The output is what you see when it runs. For the calculator, that is the tip amount and the total printed to the screen.
Once you can identify those three parts in any script, code stops being mysterious. It becomes: here is what goes in, here is the work, here is what comes out. That is all it ever is.
When It Does Not Work the First Time
Sometimes Claude's first draft does not do exactly what you wanted. That is normal and it does not mean the approach failed.
Refine the description. Tell Claude what was wrong with the previous version: "That works, but I also want it to handle the case where I split the bill between three people." Or: "The output is too technical — can you make it print a clean summary instead?"
Each iteration gets you closer. The loop is describe → generate → run → refine, not describe → perfect result. Treating iteration as part of the process — rather than a failure — is the mindset shift that makes this approach powerful.
Your Turn
Below is the BuildChallenge for this lesson. It gives you a starter version of the tip calculator — the structure is there, with three specific TODOs for you to complete. Fill in the formulas, run it, and confirm the output matches: Tip $9.00, Total $54.00.
This is a teaching scaffold — the academy shows you the code and solution for reference, but your script runs on your machine, using your own Claude session.
Once you have seen it run, you have done the thing. The loop is complete. The next lesson names what you just learned and maps where it can take you.
Continue at academy.jeremyknox.ai or explore more at jeremyknox.ai.