AI can generate a playable game quickly, but it does not remove the difficult work; it shifts that work from coding to decisions about what to make, how it should look, and whether it is enjoyable. The creator describes building two complete games without writing or reading code: Spirit Call, a top-down shooter, and Inkline Town, an ink-style town builder. Spirit Call’s detailed anime characters proved too unreliable and time-consuming for agents, so the creator changed the visual style to simple spirits that could be made and animated through a repeatable process. The lesson is that a game’s genre and style determine where human effort remains: choose a direction that lets agents handle production, or expect their limitations to become your bottleneck.
A fast first playable is only a beginning. Most later work involved playing, spotting problems, deciding what to change, and asking an agent to implement and verify the fix. In Inkline Town, for instance, fisheries made bread production pointless, an imbalance that automated bots failed to notice. Human playtesting revealed the exploit; simulations and recorded sessions helped validate a correction. The creator recommends structuring game logic with an Entity Component System and adding safeguards such as type checks, headless tests, lints, and regression tests, so agents can catch and repair many technical errors without constant supervision. These tools accelerate iteration, but they cannot judge fun, clarity, or balance. The developer’s lasting contribution is therefore not code, but taste, observation, and the choices that guide the game.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.