A two-person game studio.
One of us is a kid.
Dad runs the machines and builds the pipelines.
Zen decides whether any of it is actually fun. Nothing ships until she says so.
“The art is a big blah.”
“Bonking the blobs is too hard. Can we just delete it?”
Twenty-five years of shipping software. Builds the engines, points the agents at the boring parts, and signs the developer agreements. Then hands the build over and waits to hear what is wrong with it.
Plays every build on her iPad and files bug reports from inside the game. If the art is boring or a level drags, she says so, and it gets fixed.
A roguelike with a shovel. Clean up after an endless army of pooping cats across 18 levels, buy gear and upgrades, and take down the Supreme Cat. Built in Phaser 3, playtested every night by the Creative Director.
Zen is still deciding what we make next.
We build with AI for nearly all of it — the code, the levels, the art, and the tests that prove nothing broke overnight. What the machines cannot do is tell us whether any of it is worth playing. That judgment is the job now, and it is the one I am passing down.
Code, levels, art passes, and the test suites that prove it all still runs. Agents do that while we sleep, and they never get bored or cut a corner.
Passing tests tell you it works. They cannot tell you the second floor drags or the art is boring. Only someone playing it can do that.
Zen is not learning syntax. She is learning to play something, decide whether it is actually good, and name the problem precisely enough to fix it. That outlives every framework.