Lesson 10.1 · 45 minChoosing and scoping your next project

What you will have

The student has a one-page plan for their own next project: the user, the one job, what "done" looks like, a "not in version one" list, a build-or-buy decision with its reason, a first step and the riskiest part named.

Video

The video for this lesson is not recorded yet.

Prompts used in this lesson

Part 3 — From idea to brief

Purpose: I have an idea for a small software project for my business and I want
to turn it into a one-page brief before anything is built. Do not write code and
do not create files yet.

Context: my business is Brightside Cleaning, a home-cleaning service in one
city. I work alone on the software side and I am not a programmer. I have
already built, with Claude Code, a Next.js website on Vercel with a Supabase
database, a lead form and a private admin area. The idea: "a page where a
regular customer sees their next visits and can ask to move one."

What I want from you: interview me. Ask one question at a time, at most eight
questions, about: who exactly uses this, what they do today instead, how often
it happens, what one job the first version must do, what would make me say
"this is done", and what must not happen (data seen by the wrong person, wrong
dates, extra work for me).

Done looks like: after the questions, show me a brief with four headings:
Purpose, Context, Done looks like, Constraints. Half a page at most. Mark
anything you assumed with "ASSUMED:" so I can correct it.

Constraints: plain words, no technical terms I have not used myself. If my
answers show the idea is really two projects, say so.

Part 4 — Build, or use an existing service

Purpose: decide whether to build this or use an existing service. Still no code
and no files.

Context: the brief above.

Task: search the web for existing services that already do this job for a small
service business. For up to three, give: the name, a link to the page where you
read it, what it covers of my brief and what it does not, and where its pricing
page is. Do not quote prices or features from memory; if you could not open a
page, say so. Then answer the five build-or-buy questions for my case, and give
a recommendation: build, buy, or buy the risky part and build the rest.

Done looks like: a short table and a recommendation with its reason in three
sentences. End with the two facts I should check myself before deciding.

Part 6 — The one-page plan

Purpose: turn the cut-down idea into a one-page plan I can start this week.

Context: the brief, the build-or-buy answer, and my cuts: one user (the regular
customer), one job (knowing the next visit), one screen (an email with one
button), by hand first (I move visits myself after a Telegram message).

Done looks like: one page with these headings and nothing else:
1. The user and the one job (two sentences).
2. Done looks like (three checks a non-programmer can perform).
3. Not in version one (a list, including customer accounts).
4. Build or buy (the decision and the reason).
5. Steps: five to eight, each small enough for one sitting, each with its own
   check. The first step must be doable in under an hour.
6. The riskiest part, and how I will find out early whether it works.
7. What I need before starting (accounts, data, a decision from someone).

Constraints: if the plan does not fit on one page, cut scope, do not shrink the
wording. No step may say "and also". Mark every assumption.

Do along

Work on your own idea. Pause the video where a step says so.

1. Create a new empty folder outside your website project and start Claude Code in it. Switch to plan mode.
2. Write your idea as one sentence. If you have several, take the one that removes the most repeated work.
3. Pause after Part 3. Run the interview prompt with your own business and idea in the Context. Answer honestly; where you do not know, say "I will find out" and let it go into the brief.
4. Read every line marked "ASSUMED:" and correct it.
5. Pause after Part 4. Run the build-or-buy prompt. Open at least one link yourself. Write real costs only from pages you opened.
6. Pause after Part 5. Apply the four cuts on paper: one user, one job, one screen, by hand first. Write your "Not in version one" list.
7. Pause after Part 6. Run the plan prompt with your own cuts. Read the riskiest part, adjust, leave plan mode and save the page as plan.md.

Check your work

1. Open plan.md. Expected: it fits on one page without scrolling.
2. Read heading 1 to someone else. Expected: they can repeat who it is for and what it does.
3. Count the steps. Expected: five to eight, each with a check, and a first step you could finish in under an hour.
4. Find "Not in version one". Expected: at least three items, and nothing from that list appears in the steps.
5. Find the build-or-buy decision. Expected: a reason, and at least one link you opened yourself.

Common problems

• The plan keeps growing past one page → the idea still has two users or two jobs → tell Claude: "This must fit on one page. Propose what to move to 'Not in version one' and show me the shorter plan."
• Claude recommends building everything → the build-or-buy prompt was skipped or it answered from memory → "Search for existing services first, show links, and tell me what you could not open."

Homework

About 25 minutes, on your own, after the lesson. No other lesson depends on it.

1. A second idea, checked. Run only the build-or-buy prompt on another idea. Deliverable: a note second-idea.md in the same folder with the decision and the reason. Done when: the note contains at least one link you opened and one fact you checked on that page.
2. Show the plan to a user. Read the plan to one person who would use the result. Do not explain; ask what they expect it to do. Deliverable: three lines at the end of plan.md under "What they did not understand". Done when: the lines are there, and "Done looks like" is reworded if they misread it.
3. Book the first step. Put the first step of the plan in your calendar as a one-hour session within the next seven days. Done when: the event exists and its description contains the step and its check.

Nothing here touches your website project, so there is nothing to commit.