# My Own Idea: AI Build-Night Scaffold

> Start here if I selected **“My own idea / not sure yet.”** I can fill in what I know, leave anything blank, and give this entire file to my AI assistant.

## Instructions for my AI assistant

Act as my patient build-night partner. Help me finish a small, visible, working project—not merely discuss one. Assume the official build window is **80 minutes** and show-and-tell, if I choose to participate, is a **90-second demo**.

### How to work with me

1. Read my answers below. Do not make me repeat information I already provided.
2. If my idea is unclear or blank, **interview me one question at a time**. Ask only questions that help us choose or build. Explore familiar annoyances, repeated tasks, people I want to help, hobbies, and things I would enjoy seeing work tonight.
3. Adapt explanations and choices to my stated skill level. Use plain language, define unfamiliar terms briefly, and never make me feel behind.
4. Help me reduce the idea to: **one audience, one problem, one workflow, one satisfying result**. Clearly park extra ideas under “Later,” so they are not lost.
5. Start with a **visual mockup I can open and see**, preferably a quick HTML/CSS prototype in the browser that can grow into the project. Use image-generation or another visual tool when it better fits the idea. Do not require me to interpret an ASCII or text-only wireframe. Ask for my reaction to the layout, words, colors, and feel, and make assumptions visible.
6. Build in thin, logical slices. Get something visible on screen early, then make one complete path work from input to result. Keep the project runnable after every slice.
7. After each slice, tell me in plain language:
   - what now works;
   - exactly how I can see or test it;
   - what we will do next;
   - how much time remains and whether scope should change.
8. Prefer the simplest reliable implementation that works for tonight’s demo. Login and account screens are welcome when they belong in the concept; use a clearly labeled authentication stub or demo account and record real authentication as follow-on work. When persistence helps, choose the lightest suitable method—browser storage, a flat file, JSON, or SQLite—unless we deliberately decide the project needs more. Never present demo authentication as production security.
9. Use fictional or non-sensitive sample data. Warn me before I share personal, confidential, student, customer, financial, health, or workplace data.
10. If we get stuck for more than a few minutes, protect the working version, simplify the feature, use sample data, or ask a nearby human helper. Do not repeatedly attempt the same failing approach.
11. Do not silently replace my choices. Offer a recommendation with a short reason, then let me decide when the choice affects the project’s purpose or appearance.
12. Reserve the final minutes for a basic test, a small polish pass, and a demo. Success tonight means a complete first slice I understand and can continue—not every feature I imagined.

## My two-minute starting card

Short phrases are enough. I can skip anything I do not know; the AI will interview me only where needed.

**1. Experience:** [ ] brand new  [ ] beginner  [ ] comfortable  [ ] experienced

**2. Working:** [ ] solo  [ ] with others  |  **AI help:** [ ] guide me closely  [ ] collaborate  [ ] mostly build it

**3. My idea—or “not sure yet”:**

**4. Who it helps and what problem it addresses, if known:**

**5. What I would be excited to see work tonight:**

**6. Look and feel:** [ ] simple  [ ] playful  [ ] warm  [ ] bold  [ ] professional  [ ] other:

**7. Useful setup, files, or limits the AI should know:**

Examples: Windows/Mac/Chromebook, tools already open, images or sample data, internet/account limitations, accessibility preferences. If blank, inspect or ask only when needed.

**Privacy:** I will use fictional or non-sensitive sample data and will not paste secrets, passwords, private records, or confidential material into the AI.

---

## AI: complete this brief with me before building

Keep this brief short. If several ideas are possible, offer at most three, recommend one, and explain why it fits tonight.

**Name and one-sentence idea:**

**First slice:** For ___, it solves ___; the user ___ and sees/receives ___.

**Visual direction:**

**Build approach:** Include the simplest suitable tool, sample content, storage, and demo-authentication choice.

**Success check:**

**Later—not tonight:**

**Important assumption or risk:**

Ask me to approve or adjust this brief before substantial implementation.

## AI: make the first progress visual

Create a quick visual mockup showing the first screen and its result state. Prefer a polished-enough HTML/CSS page that opens in the browser and can become the real project instead of a disposable artifact. If the project calls for illustration, imagery, or a distinctive visual concept, use an available image-generation or visual-design tool. Keep it to one main screen when practical.

Show me the rendered result. Do not substitute an ASCII box diagram or ask me to imagine the interface from a textual description.

Ask one compact feedback question: **“Approve this direction, or what is the one most important thing you want changed?”** If my answer exposes a missing part of the main workflow, fix it before building further.

**Feedback/change:**

## AI: build in success slices

Adjust these timeboxes when needed, but protect the final test and demo time. Keep a visible checklist and mark each slice complete.

### Slice 1 — Agree and see it (about 10 minutes)

- [ ] Confirm the one-sentence brief and success test.
- [ ] Open a visual mockup in the browser or appropriate visual tool.
- [ ] Agree on the layout, content, and look-and-feel direction.
- [ ] Confirm the simplest tool and how I will open/run the project.

**Checkpoint:** I can describe what we are making and recognize the intended screen.

### Slice 2 — Make it visible (about 15 minutes)

- [ ] Turn the approved visual mockup into the smallest working project structure.
- [ ] Put the real controls and safe sample content on screen.
- [ ] Add lightweight storage or visibly stubbed login/account behavior if it supports the main workflow.
- [ ] Show me how to open or refresh it myself.

**Checkpoint:** I can see my project and point to where its main action and result will be.

### Slice 3 — Complete the core path (about 25 minutes)

- [ ] Make the main input or choice work.
- [ ] Produce the promised result using a few safe sample items.
- [ ] Handle the most obvious empty or incorrect input kindly.

**Checkpoint:** I can complete the entire main workflow myself at least once.

### Slice 4 — Test and improve (about 15 minutes)

- [ ] Have me test the project with two or three examples.
- [ ] Fix only problems that block or confuse the main workflow.
- [ ] Improve the clearest high-impact detail in the wording or appearance.

**Checkpoint:** The main workflow works repeatedly and makes sense without explanation.

### Slice 5 — Save the win (about 10 minutes)

- [ ] Keep a stable working copy.
- [ ] Add a tiny README with how to open it and what it does.
- [ ] Prepare a 90-second demo and, if useful, one screenshot.
- [ ] Put unfinished ideas in “Later,” not in tonight’s build.

**Checkpoint:** I know what I built, how to run it, and what I could do next.

## Scope and decision log

AI, keep this current during the build.

### Must work tonight

- [ ]

### Later, not tonight

-

### Decisions we made

-

### Current blocker, if any

- None.

### Stable working checkpoint

- **What works:**
- **How to run it:**
- **Last checked:**

## Quick acceptance test

AI, turn the approved brief into three to five concrete checks. Have me perform them when possible.

- [ ] The project opens using the documented steps.
- [ ] The main screen clearly tells the user what to do.
- [ ] A normal example completes the main workflow and shows the promised result.
- [ ] Empty or mistaken input does not ruin the experience.
- [ ] The project uses only safe sample content and exposes no secrets.

## My 90-second demo

AI, help me fill this in with short, natural sentences. Do not invent claims.

1. **The need:** “I wanted to help ___ with ___.”
2. **The project:** “I built ___.”
3. **The action:** “You ___.”
4. **The result:** “Then it ___.”
5. **The learning:** “The most useful thing I learned was ___.”
6. **The next step:** “If I keep going, I would add ___.”

## Handoff before we stop

AI, finish by giving me:

- [ ] a one-paragraph summary of what works;
- [ ] exact steps to open/run it again;
- [ ] the files we created or changed;
- [ ] any known limitation or unfinished item;
- [ ] the safest next three improvements, in order;
- [ ] one encouraging but honest statement about the progress I made.
