AI wrote the fighters. We kept the grudges.

The We Are So Done take

We wanted a game in which AI could help us write better tactical AI. Then AI helped build the game. Having removed a respectable amount of programming from both sides of the arrangement, we were finally free to concentrate on the important part: being annoyed that somebody else’s robots were winning.

The development conversation is almost indecently familiar. Ask us questions, one at a time. Where does the code run? How do the robots see? How do they win? Which answer feels right? The person still has opinions, constraints and taste. A large amount of the work between those decisions and something another person can use has moved elsewhere.

This is the bit that makes a programmer reach for the next project. That idea we kept postponing because it needed an engine, a viewer, accounts and a hundred boring pieces suddenly looks worth trying. We can explore something strange for the pleasure of finding out whether it works. We absolutely want more of this. We also remember when doing those hundred pieces was a substantial part of what made us valuable.

Then the game repeats the lesson at a smaller scale. We can describe a cautious fleet, an aggressive one or a strategy that should, in theory, avoid doing something visibly stupid. The assistant turns that intention into code. The arena supplies an answer that is much less interested in our intentions.

The useful feature is that there is a result to argue with. A convincing explanation of improved tactics is not a win. We can watch where the robots went, compare their decisions with what they could actually see, and ask for a change that addresses the failure. Delegation becomes much more interesting when the scoreboard does not accept prose.

None of this requires us to become indifferent. Quite the opposite. Remove enough friction from making a thing and we may become more attached to it, because now it exists. Give us another version quickly and suddenly we have time for another match. And one more adjustment. And a very reasonable suspicion that the opponent’s strategy is deeply irritating.

There is a cheerful little glimpse of our future in that loop. More of the making is delegated. More of the wanting becomes our job. We still supply the curiosity, the taste and an entirely disproportionate emotional investment in several small robots.

We have automated a lot. The grudges remain artisanal.