These browser games were developed by a friend. The early builds started with familiar arcade ideas, then moved into enemy behaviour experiments after we talked through reinforcement learning, agent movement, and how enemies can feel less random and more deliberate.
After we talked about reinforcement learning and enemy behaviour, the next build pushed the game further. The second version feels more like a deliberate prototype, with enemies and pressure becoming part of the design instead of just decoration.
The first Neon Fuse build moved the work toward a Bomberman-style arena. It is useful to keep here because it shows the earlier direction before the enemy ideas became more developed.
A Pacman-style maze chase remake that sets up the basic arcade movement language: walls, routes, pressure, and enemies that force the player to keep moving.
The game builds shown here were developed by a friend. My contribution was input and feedback around enemy behaviour, especially after sharing ideas about reinforcement learning and how enemies can create pressure through movement, timing, and decision-making.