Take a moment to think about this
In earlier lessons you learned how to write specs, how to write prompts, and how to give context. This lesson is about putting those skills into practice instead of just letting them sit in memory. There's no new teaching here — it's set up as a self-test so you can check your own work. Write out each task on paper or in a text editor, then actually run it through an AI tool.
Practice Drills
Task 1: Write a one-paragraph spec for a "to-do list app" — it needs to cover features, constraints, and edge cases, all three. Task 2: From that spec, write a coding prompt to give the AI, and check that it covers all four: role, context, constraints, and output format. Task 3: Run the prompt and see whether the output matches the spec — if it doesn't, write down exactly which point caused the mismatch, one line per issue.
Code Example
## Spec: To-do List App
### Goal
User တစ်ယောက်က task list ကို add/complete/delete လုပ်နိုင်ရမည်။
### Must have
- Add task (title text, empty မဖြစ်ရ)
- Mark complete/incomplete
- Delete task
### Edge cases
- Empty title နဲ့ submit လုပ်ရင် reject ရမည်
- Task list ဗလာဖြစ်ရင် "No tasks yet" ပြရမည်
- Title character 200 ကျော်ရင် truncate/reject ရမည်
---
## Prompt draft
Role: You are a senior frontend engineer.
Context: React + TypeScript project, function components only.
Task: Build a TodoList component per the spec above.
Constraint: no external state library, inline validation only.
Output: single .tsx file, no explanation text.You'll compare the spec, the prompt, and the AI output side by side, and come away with a list of gaps between them.Try It in 5 Minutes
Write out Task 1 and Task 2 within 5 minutes, then actually run them through an AI tool (Cursor, ChatGPT, Claude, etc.) — note down how much of the spec's coverage the output actually delivers.
A Quick Word of Caution
Don't pick a task that's too easy — throw in at least one tricky edge case, or the practice won't mean much.