Home project
Build one small, delightful webpage that recommends a short creative break — the Creative Recharge Generator.
The Creative Recharge Generator
Build one small, delightful webpage that recommends a short creative break. The experience asks three friendly questions:
- How much time do you have? — 5, 10, or 20 minutes
- What is your energy like? — low, steady, or restless
- What sounds good? — move, notice, make, or reset
It then reveals one suitable activity — for example a five-minute colour hunt, a ten-minute shape-photo challenge, a desk stretch, a tiny sound walk, or a one-sentence story prompt.
This is a good first project because it has a real flow, content decisions, simple logic, visual personality, and room for research thinking — without accounts, databases, or complicated APIs.
Brief and boundaries
Keep the first version intentionally small:
- one responsive webpage;
- an introduction, three questions, a result, and Start again;
- 8–12 recommendations stored locally in the project;
- no login, database, geolocation, personal data, external API, or health claims;
- no more than one main action per step;
- use Motion.dev only for meaningful transitions and feedback;
- use simple temporary icons or geometric placeholders — you can create final icons and illustrations later.
Step-by-step guide
Decide the experience before opening the coding tool
Write these six lines:
- Purpose: Help someone choose a realistic creative reset in under 30 seconds.
- Audience: Desk-based teammates who feel stuck or mentally tired.
- Moment: A short break during a working day.
- Tone: Friendly, calm, lightly playful, never patronising.
- Success: A new visitor reaches a recommendation without explanation.
- Boundary: This is wellbeing-inspired, not medical advice.
Then create a small assumption table:
| Assumption | How this prototype could explore it |
|---|---|
| Three questions feel quick enough | Observe whether a tester completes them without impatience |
| People understand the energy choices | Ask them to explain each choice in their own words |
| The recommendation feels achievable | Ask whether they would actually do it now and why |
Create the content first
Write 8–12 recommendation objects before asking the AI to invent the interface. Each should contain:
- title;
- duration;
- energy fit;
- activity type;
- one-sentence invitation;
- 2–3 short steps;
- a friendly completion line.
Example:
{
"title": "Five-minute colour hunt",
"duration": 5,
"energy": "steady",
"type": "notice",
"invitation": "Find one unexpected colour hiding in your surroundings.",
"steps": [
"Choose a colour you rarely notice.",
"Find three nearby objects containing it.",
"Photograph or sketch your favourite discovery."
],
"completion": "Done. You just gave your attention a different job."
}
Establish visual and interaction direction
Define the visual system in concrete language:
- warm off-white background;
- charcoal text and one fresh accent colour;
- one strong display face paired with a highly readable text face;
- generous whitespace and a narrow reading width;
- crisp borders, restrained radius, almost no shadow;
- one question per view;
- visible progress such as “2 of 3”;
- result feels like a small reveal, not a dashboard.
Create a simple flow:
Decide what happens when the user goes back, refreshes, or uses the keyboard. You may keep the first version simple, but make the decision explicit.
Prepare the project context
Create a project instruction file and include:
- the six-line product context;
- the experience flow;
- the content data;
- visual rules;
- accessibility rules;
- the approved stack;
- the instruction to use Motion.dev sparingly;
- the validation commands;
- explicit exclusions.
Ask the agent to inspect the context and return a plan before it writes code.
Build the plain vertical slice
First prompt:
Read the project instructions and recommendation data. Do not build yet.
Summarise the user flow, list assumptions or gaps, and propose a small component structure.
The first milestone is a plain but complete vertical slice: welcome, three questions,
one matched result, and Start again. No decorative motion yet.
After reviewing the plan:
Build the agreed first milestone. Use semantic HTML, keyboard-operable controls,
visible focus, and clear selected states. Keep styling simple. Do not add features
outside the brief. Run the app and production build, then summarise changed files,
validation results, and anything you could not verify.
Use the flow yourself. Commit when it works.
Add craft in three focused passes
Keep transitions quick and interruptible. With reduced motion enabled, remove movement or replace it with a simple opacity change. Commit after each pass.
Ask one person to try it
Give them the URL or your local screen. Do not explain how it works. Ask them to think aloud while finding a break suggestion. Afterward, ask only:
- What did you think this page would do?
- Was any option difficult to interpret?
- Did the recommendation fit what you selected?
- Would you actually try it? Why or why not?
Write observations before interpretations. Make one meaningful improvement based on what you learned. Do not redesign everything after one person.
Final checks and publish
Before publishing:
Deploy it using a suitable hosting service and share the link.
Role-specific lenses
Everyone completes the same project. Use your discipline to go one level deeper:
- Product designer: document the flow, selection logic, state changes, and one edge case.
- Visual designer: create a small token set and show how hierarchy adapts across viewport sizes.
- Content designer: create the recommendation content model, tone principles, and error/recovery copy.
- User researcher: define assumptions, prepare the five-minute walkthrough, and separate observations from interpretations.
These are lenses, not four separate assignments.
Definition of done
The project is complete when:
Reflection
Add five short notes to the README:
- What did AI do well?
- Where did it make an assumption you had to correct?
- Which piece of context improved the result most?
- What did your discipline notice that the AI missed?
- What would you build next — and what would you deliberately leave out?