Line 2 · the model plane
Build your own speech coach, part by part.
Eight stages. Each one is a real part of the plane. You write the code; the tests only tell you what is still broken. When every part works, you assemble your own app in your own project.
Two choices that are yours
- Build? Do you want to build a speech-practice tool? You can build and learn without entering any competition.
- Enter? Do you want to enter the Presidential SI Challenge with it? Decide by Dec 5. Building is useful either way.
If you enter: the originality rule
The Challenge says all project content must be original and created specifically for the Challenge after August 1, 2026. So:
- Start a new repository today and commit often. Your history is your proof.
- Your earlier app is your experience, not your code.
- Keep the build log from day 1 — every tool, including AI, and every bit of help. This page writes most of it for you.
Flight plan
0 / 8 partsWorkbench
- Predict what the tests will catch.
- Write your code below.
- Run the tests.
- Break it: add one test of your own.
- Why it works: one sentence.
Add a test of my own (try to break your code)
Write the inputs as a JSON array and the expected output as JSON. Example for stage 1: inputs ["one two"], expected 2.
Hints (one at a time — each one you open goes in your build log)
No tests here — you design this part. Answer on paper first, then type it in. Bring it to class: you'll choose a model together. An adult decides about any API key. Keys never go in browser code.
- Make a new project (for example
my-speech-coach) and make your first commit today. - Download your functions — the code you wrote on this page, one file:
- Download the starter kit (tests, a bare app page, missions): starter kit (.zip) Put your
my_functions.jsin it, runnode run_tests.mjs. - Record five times and fill the evidence table below. Count fillers and pauses yourself first — you are the ruler, not any other app.
- Write down the limit: five recordings help you find bugs; they do not prove the app works for other people.
The bare app.html in the kit uses Chrome's built-in speech recognition, which sends audio to Google's servers. Use it with your own voice. Before anyone else tries your app, tell them what is recorded, where it goes, and how to stop — and an adult arranges permission.
Evidence table
Your build log
Written as you go: predictions, test runs, hints opened, your own cases, your Why sentences. Add tools and help you used — be complete. Not disclosing an AI tool can disqualify a project.
Keep your work safe
Your work is saved only in this browser. Clearing browser data erases it. Download a backup often — and commit your real code to your own repository.
For your mentor
Build the parts first. If you later compare your app with any other tool (including the Studio on this site), do it after Stage 8 and say so in your write-up. Your accuracy check is your own hand count.