Nicole Greer← Back to building

Building with AI · Case study

Porchlight

A household coordination app built from a prompt.

I built Porchlight to reduce the everyday coordination friction that happens across calendars, schedules, dinner decisions, shared expenses, pet care, and repeated check-ins. Starting from an initial prompt, I used AI-assisted development to move from problem to product decisions, working software, real-world testing, and rapid iteration.

Private beta · Built for a real household problem

Claude Cowork · GitHub · Supabase · Netlify

Porchlight home view using fictional demo household data
Home turns shared signals into the next useful decision.
01

The problem

Home has a coordination layer. It just isn't designed.

Household coordination is fragmented across calendars, texts, memory, bills, work schedules, pet responsibilities, and repeated questions. The information often exists. The work is finding it, interpreting it, and knowing who needs to act.

Product principle“Use existing signals. Reduce duplicate entry. Surface the next decision that matters.”
02

What I built

One shared system for the moments that create friction.

Coordinate

Plans without the follow-up loop

Shared calendar, upcoming events, weekly schedules, and dinner timing.

Care

Leo's care, with context

Feed, walk, medication, and health tracking with timestamps.

Manage

Shared money, visible

Expenses, split amounts, and payment status in one place.

Nudge

Prompt only when action matters

Contextual prompts appear when someone needs to decide or act.

Porchlight Ahead view with fictional events
Ahead · shared calendar
Porchlight weekly schedule with fictional demo data
Weekly coordination
Porchlight dinner prompt with fictional demo data
Decision at the moment it matters
Porchlight pet care view with fictional demo data
Pet care tracking
Porchlight shared expenses with fictional demo amounts
Shared expenses
03

How I build with AI

Product judgment stays human. The implementation loop gets shorter.

I started with a problem and a prompt, then used Claude Cowork as an AI development partner. I translate real-life friction into product requirements, test the product in actual use, and iterate quickly. GitHub holds the codebase, Supabase provides the shared data layer, and Netlify handles deployment.

My role spans product manager, designer, QA, user researcher, and product owner. I am not outsourcing the product thinking to AI. I am using it to make decisions testable sooner.

01Observe friction
02Define problem
03Specify behavior
04Build with AI
05Use it
06Find friction
07Iterate
08Ship
04

Shipping velocity

Six meaningful product releases in 18 days.

  1. 01First version, local-first
  2. 02Shared sync across both phones
  3. 03GitHub auto-deploys + contextual nudges
  4. 04Expanded household scheduling + Leo tracking
  5. 05Editing + health notes
  6. 06Shared calendar + Ahead experience
“AI didn't replace product management. It shortened the distance between product thinking and product validation.”

Current status

Private product in active use.

Exploring whether Porchlight could eventually become a reusable household template for others.

Explore what else I'm building