A first-time developer made Batcula, a gothic Flappy Bird variant, with ChatGPT, Claude Code, Suno, and a little Figma cropping. The interesting part is not the “100% AI” label, but the small production workflow that turned generated concept art into a playable browser game.
The process was concrete:
- Ask ChatGPT for the game concept and visual mockups
- Give those screenshots to Claude Code as the implementation target
- Iterate for several days on levels, obstacles, villagers, scoring, and music
- Build parallax scenes by splitting a long background into sky, middle, and foreground layers
- Generate character actions as short transparent-background animation sequences
- Use Claude for deployment guidance as well as code
The author’s clearest retrospective lesson was to plan the complete game before handing it to the coding agent. Starting from a vague draft led to repeated tweaks and heavy token use; a fuller plan gave Claude Code a more stable target.
That is also what makes the project useful beyond its novelty. The tools lowered the technical entry barrier, but the result still depended on scope control, visual direction, asset decomposition, and sustained iteration—not a one-shot prompt.