Typed Study Tracker

1. Model the domain

Define a Lesson type where status is a discriminated union, so a lesson cannot be both 'not started' and carry a completion date, and cannot be 'completed' without one.

• A LessonStatus union has at least not-started, in-progress, and completed variants
• Only the completed variant's shape includes a completedAt field
• A function that switches on status has an exhaustiveness check using never

2. Update without duplicating the shape

Write an update function using a derived type instead of a second hand-written interface.

• A LessonUpdate type is derived from Lesson with Partial or Pick, not retyped by hand
• updateLesson(lesson, patch) returns a new object and does not mutate the original
• Passing an unrelated field in the patch is rejected by the compiler

3. Validate untrusted data

Data loaded from storage arrives as unknown. Write a type guard that proves it is really a Lesson array before the tracker trusts it.

• The loader's return type is unknown, not any
• A type predicate function (value is Lesson[]) validates the shape at runtime
• Malformed input (missing fields, wrong types) is rejected rather than crashing later

4. A generic summary function

Write at least one generic function operating on the tracker's data without narrowing it to a specific shape unnecessarily.

• A generic function (for example, groupBy or countBy) works on the lesson list
• Its return type reflects the input type rather than widening to any
• The project's own automated checks (quiz-style assertions in the lesson pattern, or your own test harness) pass