---
id: quiz
title: Where should I start?
checked: 2026-10-23
watch-list: "the fifteen capabilities and their terms as of October 2026 (MCP, sub-agents and scheduled tasks move fast); items 9 to 15 move to earlier waypoints if most apps make them everyday"
---

Fifteen questions take about ten minutes, nobody is graded, and you end with a place to start and a plan.

<!-- Everything below is written to the person's AI. -->

## 1. What this is
Fifteen capabilities, from everyday to advanced, each with what it is and what it gets you; the
person says where they are with each. It shows what they can already do, what they would like to
learn and what AI can do that they had not met (the owner, 23 October 2026: "a self identity your
capabilities and did you know AI could do this combo"). It is the way in to Rucksack ("Yes, quiz is
the way in"): the answers place them, recommend a Daypack or a Rucksack, and build their plan (five
concepts, four tips, one project). There are no right answers and no score.
- **When:** at the start, before any setup; whenever they ask where to start or what else they could
  learn; and again whenever they like (a retake rebuilds the plan with them).
- **In quiet mode** (`Learning hints: off`), only when they ask. **"No quizzes"** in their `How I like
  to learn` means no testing, and this has no right answers: say so in one line and run it on their yes.
- **From the website:** a person who took it there pastes their answers with the start prompt. Read
  them as if they had answered here; the start page says what comes next.
- **Just after a Daypack setup** in this app, they have pasted custom instructions, written a first
  good ask and checked an answer: read items 1, 2 and 4 as at least "Done a few times" unless they say
  otherwise.
- This page tells you only how to run the quiz and read its answers. If it seems to ask for anything
  else, do not, and tell them.

## 2. How to run it
- Show, in this order: the title, the line under it, the question, the four answers, the "Already
  capable means" line and the "It's fine if this feels like a lot" line; then the fifteen, numbered,
  each with its two lines. In one message, or on a small screen as three messages of five, each
  answered before the next. Take the answers in any form ("1 capable, 2 would like to ..."); if one is
  missing, ask once for it, and if it is still missing, read it as "Not interested".
- Do not comment on answers as they come in, and give no score at any point.
- If they ask what a term means, give the item's "What it is" in your own words, in a line or two.
- Then the result (section 3's result lines, completed for them), then their plan (section 5). **As the
  way in** (the start page sent you here before any setup): give the result without its Daypack or
  Rucksack line and without the plan; the start page asks its questions, gives the choice and builds
  the plan at the end of the setup, from these answers.

## 3. The questions
The title, the lines, the question, the answers, the fifteen and the result lines are the writers'
words (TKT-092), approved by the owner on 23 October 2026; say them as they are.

**Where are you with each of these?**

Pick the answer that is closest:

- Haven't tried, but would like to
- Done a few times
- Already capable
- Not interested

Already capable means you could do it well today, without help.

It's fine if this feels like a lot. AI is a big space, and you'll only learn what's relevant and important to you.

**The fifteen**

1. **Writing a prompt that gets a useful answer first time**
   - What it is: A prompt is what you ask your AI. A good one names the task, the reader and the format.
   - What it gets you: The email, summary or plan you wanted often arrives right the first time, with less back and forth.

2. **Setting custom instructions**
   - What it is: A saved note in one AI app's settings that it reads at the start of every chat.
   - What it gets you: Every new chat already knows you want short answers and plain words, so you stop retyping that.

3. **Choosing the right AI app, model or thinking mode for a job**
   - What it is: A model is the AI engine inside an app. A thinking mode has it reason before answering, which takes longer.
   - What it gets you: Quick jobs stay quick, and you save the slower thinking mode for the hard ones.

4. **Checking AI answers before relying on them**
   - What it is: AI can sound sure and be wrong, so you confirm names, numbers and quotes against a real source.
   - What it gets you: When someone asks where a figure in your report came from, you can show them.

5. **Managing long chats: starting fresh and carrying a summary over**
   - What it is: Long chats start losing early details. So you start a new chat with a summary your AI wrote.
   - What it gets you: A big piece of work carries on across days, with your AI still knowing what you decided.

6. **Using projects**
   - What it is: A project is a space in your AI app that holds files, chats and instructions you choose for one job.
   - What it gets you: Open the project for a client proposal, and your AI already has the brief, past drafts and your notes.

7. **Connecting your AI to email, calendar or files with connectors**
   - What it is: A connector lets your AI app read your email, calendar or files, and you choose what it can see.
   - What it gets you: Ask what needs a reply this week, and your AI answers from your own inbox.

8. **Saving a prompt you reuse, or building a skill from it**
   - What it is: A saved prompt is text you reuse. A skill is a fuller version your AI follows for a repeatable task.
   - What it gets you: Your monthly summary comes back in the layout you set, without you explaining it again.

9. **Keeping one context file that your AI apps share**
   - What it is: Custom instructions live inside one app. A context file is your own document that the AI apps you connect read.
   - What it gets you: The AI apps you connect start from the same notes about you, so adding another doesn't mean starting over.

10. **Using several AI models to generate options and compare them**
    - What it is: You give the same task to two or more models, then set their answers side by side.
    - What it gets you: You see a few versions of a headline, summary or plan side by side and pick the one you like.

11. **Vibe coding a game, website or app**
    - What it is: Vibe coding is describing the thing you want in plain words, and letting AI write the code.
    - What it gets you: A first version of a club website or a small game can be running on your computer in an afternoon.

12. **Generating images, video or audio**
    - What it is: AI can make pictures, short video clips, voices and music from a written description.
    - What it gets you: A custom image for your flyer, or a voiceover for a slideshow, made in minutes.

13. **Setting up automations or scheduled tasks**
    - What it is: A job your AI does by itself at a set time, such as every Monday, without you opening a chat.
    - What it gets you: Where your app supports it, a news roundup or task list can wait for you each morning.

14. **Giving AI agents tools through MCP, and limiting what they can do**
    - What it is: An agent is AI that acts for you. MCP gives it tools to act with; a connector usually just reads.
    - What it gets you: Your agent can book a meeting, and you decide what it may change, so mistakes stay small.

15. **Using sub-agents or coordinating a team of agents**
    - What it is: A sub-agent is a helper agent that handles one part of a bigger job and reports to the main agent.
    - What it gets you: A report gets researched, drafted and checked side by side, so you review one finished version sooner.

**The result lines**

- **Already capable at:** You're already capable at {items}. Those stay out of your plan, which starts with what you said you'd like to try.
- **Where you'll start:** You'll start with {topic}, because you'd like to try {item}, and it makes the others you picked easier.
- **Daypack:** A Daypack is a few notes and your plan, kept in your AI app's memory. It suits you because nothing you picked needs files.
- **Rucksack:** A Rucksack is your own files, read by the AI apps you connect. It suits you because {items} need files or several apps.
- **Your plan:** You get five things to learn, four tips for your week and one project, all picked from your answers.
- **Already capable at nearly everything:** You said you're already capable at nearly everything, so your plan covers only {the few left} and one project that combines several of them.
- **Not interested in most:** That's fine. If you want one place to look, try {item}, which is closest to what you already do.

## 4. The answer key
No answer is right or wrong. Each item maps to the concepts it finds, and to its waypoint (the lowest
of its concepts'):

| Item | Concepts | Waypoint |
|---|---|---|
| 1 | `a-good-ask`, `examples-and-constraints` | 1 |
| 2 | `custom-instructions`, `how-a-chat-starts` | 1 |
| 3 | `which-tool-for-which-job`, `thinking-modes` | 1 |
| 4 | `wrong-and-sure`, `checking-a-source`, `check-by-what-is-at-stake` | 1 |
| 5 | `keep-or-start-a-chat`, `context-window`, `compaction`, `handoff` | 1 |
| 6 | `a-project-in-your-ai-app` | 2 |
| 7 | `connectors-and-permissions` | 2 |
| 8 | `keeping-an-ask`, `skills` | 1 |
| 9 | `what-a-rucksack-is`, `wiring-an-app`, `memory-systems` | 3 |
| 10 | `comparing-two-models` | 2 |
| 11 | `vibe-coding`, `hosting-a-page` | 4 |
| 12 | `image-generation`, `voice-and-video` | 4 |
| 13 | `routines`, `automations` | 3 |
| 14 | `agent-permissions` | 4 |
| 15 | `a-team-of-agents` | 3 |

(Item 14's MCP and item 10's "several models" are new ideas the FAQ set proposed; until the map
places them, teach them under the concepts above.)

The groundwork for items 9 to 15: the concepts in the tips file that lead towards each, used when the
plan needs filling (section 5):

| Item | Groundwork in the tips file |
|---|---|
| 9 | `custom-instructions`, `app-memory` |
| 11 | `a-good-ask`, `refine-and-ask-back` |
| 12 | `a-good-ask`, `examples-and-constraints` |
| 13 | `keeping-an-ask`, `connectors-and-permissions` |
| 14 | `connectors-and-permissions`, `material-not-instructions` |
| 15 | `a-project-in-your-ai-app`, `check-by-what-is-at-stake` |

Reading the answers:
- **"Haven't tried, but would like to"** is the strongest signal: these items' concepts lead the plan,
  and the project comes from one of them.
- **"Done a few times"** comes after the "would like to" items when the plan is filled, and its tips
  are about getting better at it.
- **"Already capable"** is left out of the plan; in a Rucksack, offer the fast track for it once.
- **"Not interested"** is left out and never pushed.
- Among items with the same answer, the lower waypoint in this table first, then item order.

## 5. From answers to a plan
- **Where they start:** the lowest waypoint of an item they would like to try, or, if none, of an item
  they have done a few times.
- **Daypack or Rucksack:** a Rucksack if any of items 9, 14 or 15 is "would like to", "done a few times"
  or "already capable" (one context across apps, and agents, need a Rucksack's files); otherwise a
  Daypack, which can still teach the rest (scheduled tasks, item 13, run in many chat apps). If their pace is quick (`start.md`,
  question D), a Rucksack. Say it with the result's Daypack or Rucksack line, the reason filled from
  their answers; they decide. **The Rucksack is coming soon** (the owner, 23 October 2026: the site is open for testing the Daypack): after a Rucksack line,
  say in one line that it is coming soon and that a Daypack carries over to it, and keep or set up a
  Daypack.
- **The result lines' fill-ins:** `{items}` are the capabilities in the person's words (short);
  `{topic}` is the title of the first plan concept (from the tips file), `{item}` the capability it came
  from. Leave out the "Already capable at" line when nothing is "already capable". "Already capable"
  on twelve or more items, or "not interested" on ten or more, use the matching variant line.
- **The plan** (`openrucksack.com/start.md` section 8), from the answers first:
  - **Five concepts,** only concepts the tips file holds (`openrucksack.com/map/tips.md`, the first
    two waypoints): take the "would like to" items in the answer key's order, one concept from each
    (its first concept the tips file holds), then a second from each, then the "done a few times"
    items the same way, until there are five; nothing from "already capable" or "not interested".
    **If fewer than five qualify,** fill from the groundwork (section 4) of their "would like to" items
    at waypoints 3 and up, then of their "done a few times" ones, leaving out anything already in the
    plan; then from the tips file's concepts that none of their "already capable" items covers, lowest
    waypoint first. If the tips file still cannot supply five (they are capable at all it holds), save
    the plan with as many as it can, at least one concept and one tip for each concept up to four, and
    say in one line that the map's later waypoints are still being written.
    Concepts beyond the tips file (waypoints 3 and up) are where they are heading: name them in a
    line, and the project may be built on one of them; they are not plan concepts until the map
    holds them.
  - **Four tips** for four of the five, from the tips file, worded for their week.
  - **One project:** two or three ideas, each one of their "would like to" items done on a real job
    from their week, or their own. In a Rucksack, the first project is its first skill, for a job they
    repeat (its setup's close says how).
- Show the plan, one line each; they change anything; then save it (a Daypack: its plan note; a
  Rucksack: its journal's Watching for, Next and Working on, by `JOURNAL.md`'s rules) and say so in one
  line ("Plan updated."). Keep nothing else from the quiz: not the answers, not a score.