The Last Two Pieces of Software Worth Making
Most software exists to remove a constraint. A spreadsheet saves you the arithmetic. A compiler saves you the machine code. A diagnostic tool saves a doctor from guessing at a shadow on a scan. Each one takes a problem and gets you to the answer faster.
AI is unusually good at that. If a question can be stated clearly enough to compute, a model can often go further than a program a small team would build for it. Science, medicine, and engineering are full of such questions. The specialized tool starts to look like a thin window over a general intelligence, and the application becomes something that intelligence does on the way to your answer. I still care about those problems. I just don’t think they need new software from me each time.
Games don’t work like that.
Two exits.
A game is a mix of disciplines: math, engineering, art, music, psychology. Each can be solved on its own, and AI will keep getting better at solving them. But the point of a game isn’t any one of them. It’s bringing them together into something a person feels. Same ingredients, different order, different game. Two people can finish the same one and disagree about it, and neither is wrong. There is no formula to hand off.
A game is also a set of constraints. We build AI to overcome constraints. We build games to experience them. A tight jump, a small inventory, a rule you can’t break. In a tool, those are defects. In a game, they are the material, and taking them out often takes the game with it.
That makes games a good vehicle for story. Telling stories is an innate human trait. Kids assign roles in a yard long before anyone hands them a screen. A film or a book tells you a story. A game’s constraints make you live it, making the choices and paying for them.
Stories matter because they’re how we pass things down. Told well, they carry lessons and experiences we hope the next generation internalizes. A game can do that in a way that’s hard to forget, because the lesson was something you did.
I’ll always remember Sora facing Riku one last time in Hollow Bastion in Kingdom Hearts, the first time Hades waited in the snowy courtyard for Zagreus, the first Rattata I caught on Route 1 in Pokémon LeafGreen. None of those spells out a moral on screen. They stayed because I was there.
Those are the two kinds of software I still want to make. AI, because I want old limits to fall. Games, because I want to choose which limits stay, and what story gets told inside them.