How does it decide which diagonal to split on? It seems static but completely arbitrary. There were a few places where using an intersection along one diagonal would keep it open from the other route, but then it takes the other and forces you to fill the whole square.
The arrow keys are only intercepted on the page scrollbar on initial keydown. When you hold the button, the repeat inputs get through and scroll the page.
If you accidentally drag a string of cells off the right side, it'll hard-select the X option and potentially screw you up because you think left clicks still color in cells.
Here's a picture of 1.17 that illustrates what I meant be the AI just being dumb.
https://imgur.com/a/yFI9c There's not actually a logical solution to this puzzle; if the triangle played intelligently, it would take the red path and there's nothing you could do. The puzzle explicitly hinges on it being stupid for no particular reason and taking the red path, which is easy to stop. This sort of thing has already happened a few times -- sometimes where horizontally symmetrical sequences of actions will fail one way but work the other, just because in one direction the AI happens to be a moron.
AI doesn't take the red path because it tries to go nearest target using shortest way possible. Also, it'd make almost impossible to solve levels if we implemented human like decision making. However, we'll take a consider this while developing the new levels.
The game is very laggy. Also agreeing with a comment below that some of the levels are logically impossible, and you even see them as early as late world 1, but they're only solvable because the triangle isn't always that smart. It becomes a matter of finding the one path of lucky inputs that happens to manipulate the AI's seed into playing itself into a corner.
Props to your artists and just about nobody else. If I wanted to watch a fight happen with basically no significant input from me I'd watch a Lets Play of a better game.
It's interesting, but very prone to random screwage. Like in Unit Circle for instance, it's very easy to get geometrically impossible salvos within the 30-40 mark.
You had some good ideas, but the stop and wait levels were not among them. That was excruciating. Which is a shame because the puzzle in the last level was great, but the invisible platforms just made it tedious as hell.
 "Neo Angle"){kind=icon}
 "SWITCH or NOT?"){kind=icon}
 "The Escape Exam"){kind=icon}
 "Pixoji"){kind=icon}
 "Hexa turn"){kind=icon}
 "Tesla8"){kind=icon}
 "Save Her!"){kind=icon}
 "Song for a bird"){kind=icon}
 "Forgotten Hill: Surgery"){kind=icon}
 "Heroes and Titans"){kind=icon}
 "Beam Ball"){kind=icon}
 "Unstable"){kind=icon}
 "Fill Tracks"){kind=icon}
 "Castle Wars 2.5"){kind=icon}
 "The Book of Ethan"){kind=icon}
 "Save To Kill"){kind=icon}
AI doesn't take the red path because it tries to go nearest target using shortest way possible. Also, it'd make almost impossible to solve levels if we implemented human like decision making. However, we'll take a consider this while developing the new levels.