Cards with Cats

Cards with Cats

Cards with Cats

Cards with Cats

AI Development

AI Development

AI Development

Overview

Overview

Cards with Cats is a "Cards Against Humanity"-style party game where you play against a cast of AI cats who deal, judge, trash-talk, and crown winners in their own distinct voices. It runs on a single device for hot-seat night or across phones and a TV for Party Mode. But the real project wasn't the game — it was the question behind it: how far can one non-traditional developer get by orchestrating today's AI tools instead of a full team? The answer turned into a fully self-hosted, production-grade app with nearly 30,000 cards, 48 custom-illustrated decks, and cats with opinions.

Cards with Cats is a "Cards Against Humanity"-style party game where you play against a cast of AI cats who deal, judge, trash-talk, and crown winners in their own distinct voices. It runs on a single device for hot-seat night or across phones and a TV for Party Mode. But the real project wasn't the game — it was the question behind it: how far can one non-traditional developer get by orchestrating today's AI tools instead of a full team? The answer turned into a fully self-hosted, production-grade app with nearly 30,000 cards, 48 custom-illustrated decks, and cats with opinions.

Tech Stack

MidJourney

AI

Eleven Labs

AI

ChatGPT

AI

Claude Code

AI

Higgsfield

AI

Docker

Development

React

Development

Tech Stack

MidJourney

AI

Eleven Labs

AI

ChatGPT

AI

Claude Code

AI

Higgsfield

AI

Docker

Development

React

Development

Tech Stack

MidJourney

AI

Eleven Labs

AI

ChatGPT

AI

Claude Code

AI

Higgsfield

AI

Docker

Development

React

Development

Tech Stack

MidJourney

AI

Eleven Labs

AI

ChatGPT

AI

Claude Code

AI

Higgsfield

AI

Docker

Development

React

Development

Started

Started

2026

Play the Game

Play the Game

The Project Process

The Project Process

Exploring the Challenge
Shipping a game like this normally takes a studio: engineers, illustrators, copywriters, a voice cast, and someone to keep the servers alive. I had none of those — just a clear picture of what I wanted and a stack of AI tools. Could that be enough to build something genuinely polished, not a prototype? And could it handle the hard parts on its own terms — real multiplayer, a self-hosted backend, and a content system that's safe for family game night and able to get delightfully wild after the kids go to bed?

The Design Approach
I treated AI as a team I directed rather than a button I pushed. Claude Code acted as the engineering lead — a React + Vite front end, a Fastify/SQLite API, multi-device rooms, and Docker deployment to my own TrueNAS server — alongside a custom pipeline that generates and safety-filters thousands of themed cards. ChatGPT was my sounding board for ideation and copy. MidJourney gave the cat cast and table themes their look; Higgsfield illustrated a bespoke banner and matching card back for every single deck; and ElevenLabs gave the cat judges their voices, so a verdict is actually spoken in character. The thorniest design problem was content: a dual Family / Adult system with an age gate and layered guardrails, so the same engine can host a wholesome round or an unapologetically R-rated one — safely and on purpose.

The End Results
A living, self-hosted product: 48 decks, each individually illustrated, spanning everything from Bluey to a Jane Austen "Wicked Edition." Nearly 30,000 cards. AI cat judges with real personalities who react to the exact card that won — out loud. Single-device hot-seat and multi-device Party Mode. All built, deployed, and running, by one person directing a toolchain.

The Project Process

Exploring the Challenge
Shipping a game like this normally takes a studio: engineers, illustrators, copywriters, a voice cast, and someone to keep the servers alive. I had none of those — just a clear picture of what I wanted and a stack of AI tools. Could that be enough to build something genuinely polished, not a prototype? And could it handle the hard parts on its own terms — real multiplayer, a self-hosted backend, and a content system that's safe for family game night and able to get delightfully wild after the kids go to bed?

The Design Approach
I treated AI as a team I directed rather than a button I pushed. Claude Code acted as the engineering lead — a React + Vite front end, a Fastify/SQLite API, multi-device rooms, and Docker deployment to my own TrueNAS server — alongside a custom pipeline that generates and safety-filters thousands of themed cards. ChatGPT was my sounding board for ideation and copy. MidJourney gave the cat cast and table themes their look; Higgsfield illustrated a bespoke banner and matching card back for every single deck; and ElevenLabs gave the cat judges their voices, so a verdict is actually spoken in character. The thorniest design problem was content: a dual Family / Adult system with an age gate and layered guardrails, so the same engine can host a wholesome round or an unapologetically R-rated one — safely and on purpose.

The End Results
A living, self-hosted product: 48 decks, each individually illustrated, spanning everything from Bluey to a Jane Austen "Wicked Edition." Nearly 30,000 cards. AI cat judges with real personalities who react to the exact card that won — out loud. Single-device hot-seat and multi-device Party Mode. All built, deployed, and running, by one person directing a toolchain.

A human note on the tools

This was, honestly, amazing to build. These tools are profoundly empowering: they let an individual make almost anything they can picture, and that's worth celebrating. But I want to be clear about what I believe. AI is not a replacement for real people — it's at its best amplifying human creativity, not standing in for it. It should not become a quiet excuse for organizations to cut staff purely to save money; the goal should be helping people do more, not erasing the people. And it isn't free: the energy and water that AI and its data centers consume carry a real environmental cost we shouldn't look away from. I built this with AI, eyes open — grateful for the leverage, and unwilling to pretend it's magic without consequences.

A human note on the tools

This was, honestly, amazing to build. These tools are profoundly empowering: they let an individual make almost anything they can picture, and that's worth celebrating. But I want to be clear about what I believe. AI is not a replacement for real people — it's at its best amplifying human creativity, not standing in for it. It should not become a quiet excuse for organizations to cut staff purely to save money; the goal should be helping people do more, not erasing the people. And it isn't free: the energy and water that AI and its data centers consume carry a real environmental cost we shouldn't look away from. I built this with AI, eyes open — grateful for the leverage, and unwilling to pretend it's magic without consequences.

The Project Process

The Project Process

Exploring the Challenge
Shipping a game like this normally takes a studio: engineers, illustrators, copywriters, a voice cast, and someone to keep the servers alive. I had none of those — just a clear picture of what I wanted and a stack of AI tools. Could that be enough to build something genuinely polished, not a prototype? And could it handle the hard parts on its own terms — real multiplayer, a self-hosted backend, and a content system that's safe for family game night and able to get delightfully wild after the kids go to bed?

The Design Approach
I treated AI as a team I directed rather than a button I pushed. Claude Code acted as the engineering lead — a React + Vite front end, a Fastify/SQLite API, multi-device rooms, and Docker deployment to my own TrueNAS server — alongside a custom pipeline that generates and safety-filters thousands of themed cards. ChatGPT was my sounding board for ideation and copy. MidJourney gave the cat cast and table themes their look; Higgsfield illustrated a bespoke banner and matching card back for every single deck; and ElevenLabs gave the cat judges their voices, so a verdict is actually spoken in character. The thorniest design problem was content: a dual Family / Adult system with an age gate and layered guardrails, so the same engine can host a wholesome round or an unapologetically R-rated one — safely and on purpose.

The End Results
A living, self-hosted product: 48 decks, each individually illustrated, spanning everything from Bluey to a Jane Austen "Wicked Edition." Nearly 30,000 cards. AI cat judges with real personalities who react to the exact card that won — out loud. Single-device hot-seat and multi-device Party Mode. All built, deployed, and running, by one person directing a toolchain.

A human note on the tools

This was, honestly, amazing to build. These tools are profoundly empowering: they let an individual make almost anything they can picture, and that's worth celebrating. But I want to be clear about what I believe. AI is not a replacement for real people — it's at its best amplifying human creativity, not standing in for it. It should not become a quiet excuse for organizations to cut staff purely to save money; the goal should be helping people do more, not erasing the people. And it isn't free: the energy and water that AI and its data centers consume carry a real environmental cost we shouldn't look away from. I built this with AI, eyes open — grateful for the leverage, and unwilling to pretend it's magic without consequences.

Next Project

Next Project

Next Project

->

->

Devon Scott Burch © 2026

Devon Scott Burch © 2026

Create a free website with Framer, the website builder loved by startups, designers and agencies.