How a 91-Year-Old Vibe-Coded a Complex Church App with Claude and Replit

By: John Blackman Host: Claire Vo Source: How I AI (ChatPRD) Type: podcast

Summary

How I AI episode with John Blackman, a 91-year-old retired electrical engineer, and his grandson Brandon. John built a sophisticated multi-tenant event management application for his church’s “Impact Weekends” (community outreach events providing free services) with no prior software development experience. Two-part workflow: (1) Claude as AI product manager — fed a detailed Word document describing the user journey, Claude generated professional-grade user stories for multiple roles (system admin, local church admin, pastor, participant), designed multi-tenant architecture with role-based access control, proposed an MVP with phased rollout plan, and outlined non-functional requirements for security and privacy; (2) Replit agent as AI developer — pasted entire Claude requirements blueprint into Replit agent, which immediately generated file structure, database schema, and frontend/backend code; iterative vibe-coding cycle of prompting, observing, testing, and providing feedback. Final application features: secure admin dashboard with multi-tenant access, event and service management with toggle controls, automated Excel reporting suite (demographics, service usage, shopping lists), VIN lookup via OpenAI API for oil change shopping lists, QR code participant registration, digital signature waivers, personalized “Impact Passport” PDFs emailed via SendGrid. Built for ~$350 in a few weeks of late-night sessions vs. estimated six months with a traditional team.

Key Ideas Extracted

Claude as AI product manager: Feed a detailed Word document of your vision and get back professional-grade user stories, multi-tenant architecture, MVP definition, phased rollout, and non-functional requirements
Replit agent as AI developer: Paste the entire Claude blueprint into Replit and the agent builds file structure, database schema, and full-stack code — John described watching in amazement at the speed
Two-tool workflow: plan then build: Use one AI (Claude) for product planning and another (Replit) for implementation — clear handoff via copy-pasting the requirements document
Iterative vibe-coding feedback loop: Prompt → observe → test → feedback; know when to say “Wait” or “Stop” when the agent goes down a “rabbit trail”
$350 vs. 6 months: A non-developer built a complex multi-tenant app with API integrations for roughly $350 in a few weeks, replacing what would have been a six-month professional development effort
Lifelong learning mindset: John compares AI adoption to learning AutoCAD in the 1980s — those who embraced new tools extended their careers and continued having impact
Sophisticated features from conversational prompts: Multi-tenant access control, VIN lookup API integration, SendGrid email delivery, digital signatures, Excel report generation — all built through natural language interaction
AI empowers community problem-solving: John saw a real problem (paper-based event registration) and had the tools to solve it himself, creating positive community impact without waiting for budget or developers

Notes

Published Jun 23, 2025 on How I AI (ChatPRD). ~8 min read.
Sponsors: WorkOS, Orkes
John Blackman background: 91-year-old retired electrical engineer; first in his department to learn AutoCAD in the 1980s; helped launch Google Fiber
Tools: Claude (planning), Replit agent (coding), OpenAI API (VIN lookup), SendGrid (email), QR codes
App: Church “Impact Weekends” event management — community outreach providing free haircuts, dental care, food, oil changes
Feature highlight: VIN lookup generates precise oil type/amount and filter shopping lists for volunteer mechanics
Two companion workflow guides published Jan 8, 2026
Cross-references: Vibe coding, non-technical builders, AI democratizing software development

Raw Content

Re-scraped from ChatPRD 2026-02-16. Full article content captured in Summary and Key Ideas above.