Josh Rutz · Technical direction
AI in games
Where it belongs, and where it doesn't.
A working point of view, built from shipping with it.
Scroll
The frame

AI is the next abstraction, not the end of the craft.

C++ pushed assembly down to a footnote. It didn't kill the craft of engineering, it moved it up a level. AI does the same thing to making games. The low-level work gets cheaper, and the taste, the structure, and the judgment matter more, not less.

This isn't theory

I already build this way.

Zipix is a live micro-game studio I built solo, AI in the loop end to end, before hiring anyone.

0
Games shipped on one shared engine contract
0
Agent playbooks, one per discipline, codified and reused
0
Weeks from seed to live in production
The point of view below is where I already work, not where I'm guessing.
Where AI belongs

On the boilerplate, so people spend their budget on what needs a human.

Production leverage
Pipelines, tooling, content scaffolding. The work that's necessary but not creative.
Velocity
One operator producing the output of a team, with a human checkpoint at every stage.
Coverage
Tests and gates that keep quality from drifting while the team moves fast.
The line

AI lowers the cost of making. It must not lower the cost of caring.

A player gives maybe a tenth of the effort to play a thing that I spent days making. The moment a thing is cheaper to generate than it is to play, the player feels it, and they're right to. AI is leverage on effort, never a substitute for it.

Where AI doesn't belong

Not on the part that makes a game worth the minute.

The emotional spine
What a game is reaching for is a human call. Generated feeling reads as generated.
The core loop's taste
What's fun, what's fair, what's worth the player's time. That's judgment, not generation.
Player trust
Safety, fairness, and honesty with players are guardrails AI sits inside, never owns.
How I'd run it

Amplify taste, with a tested path back.

Content velocity
AI fills the pipeline. Humans pick what ships.
Trust and safety
Generation runs inside guardrails, reviewed, with kill switches.
Live-ops ethic
Degradation pages me, not the players.
Hiring
AI opens the pool to good ideas and good practices over knowledge moats. I'd hire for taste.
The point of view

AI changes who can make games. It doesn't change what makes one worth playing.

I'd use it to put a small team's leverage in every maker's hands, and spend the time it frees on the decisions only a human should make.