Choose Your Topic
A step-by-step framework for deciding what to build as your parallel project — done in one sitting, before Week 3 Monday.
You have until the end of Week 2 to decide what you're building. This page walks you through the decision in four steps. Do it in one sitting — it should take under 30 minutes.
The One Question
What is a problem I deal with regularly that no software solves well — or solves with clunky software I hate?
That is the only question that matters. Your parallel project should be the answer to it.
If you are a CA, you have a lifetime of these. Compliance tracking, fee registers, client documents, meeting notes — you live with friction that most software developers have never experienced. That is your advantage: you understand the problem deeply before writing a single prompt.
If you have a personal problem that fits (a family business inventory, a habit tracker, a household tool), that works too — as long as you understand the problem and will actually use the result.
Step 1 — List Your Problems
Spend 10 minutes writing down every "I wish there was software for this" moment from the last month. Do not filter yet.
Starter questions to jog your memory:
- What do you track in Excel that Excel handles badly?
- What do you look up in WhatsApp, notebooks, or folders because there is no better system?
- What deadline or task have you forgotten at least once in the last year?
- What takes 20 minutes every week that should take 2 minutes?
- Where do you do copy-paste work that should be automatic?
- What do your clients ask you for that takes effort to look up?
Write the answers in any order. Aim for 5–10 items. Messy is fine.
Step 2 — Score Each Idea
For each item from Step 1, answer these four questions. Score 1 point for each "yes":
| Question | Yes | No |
|---|---|---|
| Do I personally live with this problem at least once a week? | 1 | 0 |
| Could I explain the full problem in two sentences right now? | 1 | 0 |
| Does it involve data I could put in 2–3 database tables? | 1 | 0 |
| Would I actually use a working version of this app? | 1 | 0 |
Scoring:
- 4/4 — build this one. This is your project.
- 3/4 — strong candidate. Identify which question was "no" and ask if you can work around it.
- 2/4 or below — move to the next idea. Do not force it.
If multiple ideas score 4/4, pick the one where the problem costs you more time or money. That one has more motivation fuel.
Step 3 — Scope Check
A parallel project built in 2–3 hours per week across 8 weeks has a strict size limit. Use this filter before committing to your idea.
It fits if:
- The core data lives in 2–3 database tables
- There is one primary user (you — or one other specific person)
- The core loop is 3–4 steps (e.g., create client → add task → view dashboard → mark done)
- You can describe the minimum useful version in one sentence
It does not fit if:
- It requires 5+ tables to be useful at all
- Multiple people need different roles from Day 1
- It requires external integrations (SMS, Razorpay, email, government APIs) before it is useful
- You cannot describe the core loop without saying "it depends"
Scope rescue: If your idea is too large, find the smallest slice of it that is useful on its own.
"A billing and client management and compliance and document platform" → "A fee register — what each client owes me, what's been paid, what's outstanding." Start there. Everything else can come after Demo Day.
Step 4 — Write the 3-Sentence Pitch
Before you can build it, you must be able to describe it in exactly three sentences. If you cannot write these sentences cleanly, the project is not yet well-scoped enough to build.
The template:
- What it does: "A tool to [core action] for [who or what]."
- Who uses it: "Just me" or "Me and [one specific person]."
- What the main data is: "The main data is [table 1] and [table 2]."
Examples:
- A tool to track which clients have filed their monthly GST returns and flag who is overdue.
- Just me — I'm the only user.
- The main data is clients (with GSTIN) and filing records (month, status, filing date).
- A tool to record what was discussed in each client meeting and what the next steps are.
- Just me.
- The main data is clients and meeting notes (date, summary, next steps, follow-up date).
- A simple inventory tracker for my family's hardware shop — items, quantities, and restocking alerts.
- Me and my brother — two users, same view, no roles needed yet.
- The main data is products (name, unit, low-stock threshold) and stock adjustments (date, quantity, reason).
- A fee register for my CA practice — what each client owes, what has been paid, and what is outstanding.
- Just me.
- The main data is clients and fee entries (service, amount, invoice date, paid date, outstanding).
If you can write those three sentences, you have a project. Bring them to Week 3 Monday — Claude Code will scaffold the whole thing in one prompt.
What If I'm Still Stuck?
If you reach the end of Step 3 and nothing feels right, use the CA compliance tracker as your project. It is a ready-made idea with a complete worked example covering every week's homework milestone.
The CA compliance tracker — tracking GSTR-1, GSTR-3B, ITR deadlines per client — has a full Schema & Setup page, a Week-by-Week Build Guide, and copy-paste Bootstrap Prompts. If you choose it, you can follow those three pages directly without adapting anything.
Choosing quickly and building is better than spending two weeks deciding. The goal is to ship — not to find the perfect idea.
Before Week 3 Monday
Once you have your three sentences, do two things before the Monday session:
- Share your three sentences — drop them in the group before Sunday night. This is your public commitment. Saying it to the group makes it real.
- Think about your two main tables — what would you name them? What are the most important columns? You do not need to write any SQL — just think through what data you will store. Ten minutes of thought here saves 30 minutes of confusion during scaffolding.
Claude Code will build the entire project skeleton in the first 30 minutes of your Week 3 homework session.