Skip to main content
Harmanpreet Singh
All work

Independent project: Interactive learning lab

LaunchHelp: from an idea to a plan you can build

An interactive lab that turns a product idea into a buildable plan across thirteen phases, from the idea brief and problem tests to an API contract, a test matrix and a launch checklist, with a 38-chapter handbook alongside. It runs entirely in the browser and keeps assumptions apart from evidence.

What it is

Each phase has a short lesson, one interactive exercise, a saved versioned artifact and an explicit completion checklist. The artifacts export as a ZIP of markdown files and an OpenAPI draft, and an unfinished phase is listed as pending.

It is careful about what it will not claim: an idea is not called validated without evidence the user recorded, a test plan says it was not run, and nothing is called deployed. Work is stored only in the browser, so a backup download matters, and AI drafting, collaboration and a backend are planned rather than built.

My part

I designed and built the app, the planning logic, the export and the handbook.

Independent project, a sibling of MLForge and SystemForge.

Stack

  • React 18
  • TypeScript
  • Vite
  • Vitest
  • Playwright

What is inside

A static React app over a plain TypeScript core with no React in it.

  1. 01

    Revisions and conflicts

    Every autosave is a revision. A change made in another tab opens a conflict dialog that keeps both versions, and history shows diffs.

  2. 02

    Approval and traceability

    Approving pins a revision, and exports use approved revisions only. Changing something other work depends on marks the dependents as needing review, and nothing is rewritten silently.

  3. 03

    Generators and export

    Scope, architecture, schema, API, tests, board and checklist generators feed a store-only ZIP writer. The API file is held back, with a reason, if a state-changing endpoint lacks authorization.

  4. 04

    Typed content

    Phases, lessons and the 38 handbook chapters are typed records, and tests enforce content rules such as source dates, no placeholder text and no claims that tests passed or something deployed.