How to use Claude to build a parent learning platform
The author used Gemini Pro for architecture and Claude Code for implementation, enforcing Vitest tests before deploying My DigitAlly to serve 200+ active users.
news.ycombinator.com
Suggested prompt
Suggested prompt written by Agents In Use, not the source author. Replace bracketed inputs and review tool permissions first.
Implement [parent education feature] in [repository] using the architectural decisions in [context document]. Write failing Vitest tests first, implement the feature, and report the test results. Keep [architecture assistant] responsible for architecture and require passing tests before deployment.
Reach people exploring AI agents
See placements →Evidence at a glance
Sourced account- Source author
- kanchanepally
- Original account
- Read the source ↗
- Published
- Source link checked
- Reported result
- The author reports reaching production in four weekends and running a stable system serving 200+ active users.
- Human involvement
- The author required every AI-generated feature to follow a Vitest TDD cycle and withheld production deployment when tests failed.
A linked account supports attribution. It does not mean we independently reproduced its result.
Report an issue with this case
Open correction emailOpens your email app with the case link included. Nothing is sent automatically.
How the agent did it
- Use Gemini Pro for architecture and product prioritization.
- Use Claude Code to implement cron jobs and refactor API patterns.
- Write the test suite with Claude Code.
The author reports taking My DigitAlly, an educational platform for parents, from an initial idea to production over four weekends while maintaining a full-time role. The application includes Google OAuth and email/password authentication, weekly tips sent by a node-cron system, a JSON-driven lesson flow, and authenticated server-side PDF generation for parent checklists.
The development approach assigned different responsibilities to different assistants. Gemini Pro handled architectural decisions and product prioritization, including choices around React/Vite, Node/Express, and PostgreSQL/Prisma. Claude Code handled implementation work, specifically cron jobs, API-pattern refactoring, and the test suite.
The author describes a strict test-driven development cycle using Vitest for every AI-generated feature. Passing tests was a requirement for sending code to production. They also maintained context documents and boundaries between the assistants' responsibilities, which they credit with keeping the codebase consistent.
The stated result is a stable production system serving more than two hundred active users, with weekly tips going to more than two hundred subscribers. Those are the author's reported outcomes, rather than independently verified usage figures. Who it's for: developers who want separate architectural and implementation roles for AI assistants, with an explicit testing requirement before deploying an educational application.
Prompt
Suggested prompt (written by Agents In Use, not the author's)
Source: Hacker News · kanchanepally
Frequently asked questions
What did Claude Code implement?
The post specifically names cron jobs, API-pattern refactoring, and writing the test suite.
What was the deployment gate?
Every AI-generated feature followed a strict Vitest test-driven development cycle; code did not go to production if its tests failed.
More details
Remaining steps
- Require passing Vitest tests before production deployment.
About this account
Claude Code implemented application features and tests, while Gemini Pro handled architecture and prioritization within separately maintained contexts.
Platform: Hacker News. Attribution: https://news.ycombinator.com/item?id=46436324. Prompt: Suggested (written by Agents In Use, not the author's). Sourced by the Agents In Use pipeline.
Credit: kanchanepally on Hacker News
Source published
Added to Agents In Use
Source checked
Tools and topics
Claude Code · Gemini Pro · Vitest · node-cron · PostgreSQL · Prisma
AI development
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.