Skip to main content

Morning Brew

Unity Engine | Bezi AI | Team Project

About The Game​

Morning Brew is a cozy coffee simulation game developed for the Bezi Jam, a monthly game jam focused on using the Bezi AI agent for Unity. The theme was "Cozy Games," and the objective was to create a relaxing experience where players can brew a cup of coffee, create latte art, and enjoy their creation on a rainy morning.

The core loop involves pulling an espresso shot, using a drawing mechanic to create latte art, and enjoying a cup of coffee. The game was built to test the capabilities of AI-assisted development workflows.

Media Showcase​

The espresso machine interface where players pull a shot of coffee.
The espresso machine interface where players pull a shot of coffee.
The espresso machine interface where players pull a shot of coffee.Zoomed out espresso machine sceneThe latte art mechanic allowing players to draw on the coffee foam.Enjoying your cup of coffee on a cozy morningClock ticking scene

My Role & Technical Contributions​

I worked as the lead developer and designer in a team with a composer and an artist. My primary focus was integrating the Bezi AI tool into the workflow to accelerate development.

  • AI-Assisted Development: I utilized Bezi, an AI agent for Unity, to generate scripts and shaders. This included setting up the project structure and attempting to generate complex systems like the latte art shader.
  • Latte Art Mechanic: I implemented a drawing system using Render Textures. This allows players to draw directly onto the coffee sprite. While Bezi helped generate the initial boilerplate, I had to refine the coordinate mapping and input detection to make the drawing feel natural.
  • Game Feel & Polish: I focused on the "cozy" aesthetic, implementing camera movements, transition effects (like a fading clock), and integrating sound effects to enhance the atmosphere.

Lessons Learned​

  • AI as a Copilot, Not a Captain: Working with Bezi highlighted that while AI can speed up boilerplate code, it cannot replace a deep understanding of the systems being built. I learned that researching the underlying technology (like Render Textures) is crucial before asking an AI to implement it, otherwise debugging becomes significantly harder.