How to use Claude to ship mobile puzzle games
Omar Alhaddad shipped two mobile puzzle games with Claude, iterated on shaders and used custom skills for achievement icons and store screenshots he reviewed.
Original prompt
Shared by the source author. See source notes for context.
check @achievements.json file, for each item there, use /generate-image to create an icon,
Reach people exploring AI agents
See placements →Evidence at a glance
Sourced account- Source author
- Omar Alhaddad
- Original account
- Read the source ↗
- Published
- Source link checked
- Reported result
- Block Orbit and Gridrise shipped to linked app stores, with generated gameplay, shaders and publishing assets reviewed and iterated by Alhaddad.
- Human involvement
- Alhaddad set the game designs, maintained architecture notes, described visual defects and reviewed code because shader fixes could introduce subtle bugs.
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
- Alhaddad specified gameplay and maintained architecture and design notes.
- Claude implemented mechanics and iterated on Metal and Three.js rendering.
- Custom skills generated achievement icons and store screenshots for human review.
Omar Alhaddad shipped Block Orbit and Gridrise after using Claude Code across gameplay, rendering and publishing work. The single retrospective covers both games: a block puzzle on a rotating cylinder and a tower-placement puzzle whose clues depend on edge visibility.
Claude implemented mechanics, interface work, save systems, level parsing and platform integrations. Rendering used different stacks: Swift and Metal for Block Orbit, and React Native with Three.js for Gridrise. Alhaddad describes substantial iteration on shaders and game feel rather than an unreviewed first generation.
Publishing chores became part of the agent workflow. A custom image skill called Gemini's image API to produce achievement icons, inspected results against requirements and retried unsuitable images. Another skill generated store screenshots in a Next.js project; Alhaddad requested added iPad support when the initial version lacked it.
The source links released games and describes the human work that remained. Alhaddad kept architecture and design notes, supplied precise visual descriptions and reviewed code because rendering corrections could introduce subtle defects. Claude helped implement the chosen games and package their assets, while their design direction and release judgment stayed with the developer.
Prompt
Original prompt
Source: Blog · Omar Alhaddad
Frequently asked questions
Were the games released?
The retrospective says both shipped and links their store pages.
How were achievement images generated?
A Claude skill called Gemini's image API and checked images against requirements.
Why did rendering need supervision?
The author had to describe visual defects and review fixes for subtle new bugs.
More details
About this account
Claude implemented mechanics, interface work, save systems, level parsing and platform integrations. Rendering used different stacks: Swift and Metal for Block Orbit, and React Native with Three.js for Gridrise. Alhaddad describes substantial iteration on shaders and game feel rather than an unreviewed first generation.
Primary retrospective opened. Two games are linked to stores in one canonical account, kept as one entry. Original prompt excerpt is from achievement generation. Release and productivity claims are author's reports. Original prompt excerpt retained from the source.
Credit: Omar Alhaddad on Blog
Source published
Added to Agents In Use
Source checked
Tools and topics
Claude Code · Metal · Swift · React Native · Gemini image API
Mobile games · Publishing assets
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.