MSB-341_Shulin_Repo

PengPal

Travel China like you have a friend there: PengPal connects first-time foreign travelers in China with young, English-speaking Chinese locals who help with transportation, payments, food, and local experiences.

Live prototype: https://shulinzzz.github.io/MSB-341_Shulin_Repo/product/pengpal/ Previous project (Sprints 1–2): MCAT Prep Tracker / Gapmap at https://shulinzzz.github.io/MSB-341_Shulin_Repo/product/mcat/ Side project: To-do list app with Supabase backend (spec 003) at https://msb-341shulinrepo-todo.vercel.app Built by: Shulin Zhang, MSB 341 Product Management, BYU

Context

Fill this in during Sprint 1 and keep it current. Every sprint is read against it.

• What I am building: PengPal, a service that connects first-time foreign travelers in China with young, English-speaking Chinese locals who help them navigate transportation, payments, food, and authentic local experiences. Pilot cities: Chongqing + Shanghai (decisions/003). Right now it is a clickable prototype (specs/004); the first trips are meant to be hand-matched before building real software.
• Who it is for: First-time foreign travelers to China, and the English-speaking local hosts who guide them.
• My role: Product lead, PM, and builder on a team.
• My user: People planning a first-time trip to China in the next year.

What changed (Oct 14, Sprint 3): Switched from the MCAT Prep Tracker / Gapmap to PengPal. The Sprint 2 tests showed Gapmap would need accurate assessments, flexible scheduling, and progress tracking to be useful, and would compete with established MCAT platforms. PengPal can be validated faster with real, hand-arranged trips (see sprints/sprint-2-review.md, retro).

If your situation changes, revise this and note what changed. That is normal; a silent mismatch between this file and your work is not.

Where the work lives

• Repos: ShulinZZZ/MSB-341_Shulin_Repo, public, shared with sdmurff and the TA
• Live:
  • MCAT Prep Tracker: https://shulinzzz.github.io/MSB-341_Shulin_Repo/product/mcat/
  • PengPal clickable prototype: https://shulinzzz.github.io/MSB-341_Shulin_Repo/product/pengpal/
  • Gapmap with Knowledge Tree (prototype used in Sprint 2 tests): https://claude.ai/artifact/42hitm4hF1kYmTd2xPHfnx
  • To-do app (side project): https://msb-341shulinrepo-todo.vercel.app
• Docs:
  • Sprint 2 interview and test notes: discovery/interviews/ (002–012, plus the Word notes and summaries)
  • PengPal pitch deck: discovery/PengPal/PengPal-Big-Idea-Pitch-Guided-Travel-Final.pptx
  • PengPal 90-second pitch video: discovery/PengPal/PengPal Pitch Video.mp4

What is in this repo

Folder	What lives here
sprints/	One plan and one review per sprint
discovery/	Interviews, personas, what you learned about your user
design/	Flows, screens, usability test notes
product/	The build itself
specs/	One spec per feature, written before building it
gtm/	Launch, channels, copy, experiments
metrics/	What you measure and what it says
decisions/	Numbered records of what you decided and why

Non-code work belongs here too. An interview, a pricing model, a landing page draft, and a usability finding are all artifacts, and they get committed like anything else.

If you build an AI feature, put its eval set in product/evals/. A test set is how you know whether a change to a prompt helped or hurt.

Running it

Open the PengPal prototype: https://shulinzzz.github.io/MSB-341_Shulin_Repo/product/pengpal/ (the MCAT tracker is still at /product/mcat/).

No build step or install. To run locally, serve the repo root and open /product/pengpal/:

npx serve .        # or: python -m http.server

Data is stored in your browser’s localStorage, so each browser has its own data.

Sprints

Each sprint:

/sprint-plan     # day one, then commit the plan
# ...build...
/sprint-review   # last day, then commit the report and write your retro

Ground rules

• Spec before build. For anything non-trivial, the spec’s commit should predate the feature’s commits.
• Decisions get recorded. When you make a real choice, write it in decisions/ with the alternatives you rejected.
• No real customer contact details anywhere in this repo. Anonymize people in interview notes: “dental office manager, Provo” rather than a name and an email.
• Keep CLAUDE.md current. It is what your agent knows about your work. Stale context produces bad output.