The port that argued against itself

A working game was ported to a second engine as a trial, one gated step at a time, with captures of the same frame in both.

It reached parity. That was the point of the trial, and it would have been the moment to declare victory.

The verdict recommended staying where it was. The new engine spent 1.6 to 2.1 times the GPU per frame. The fast gate went from 35 seconds to about four minutes. The game code grew by 31 percent. The old engine had produced about 50 written-down traps across three milestones. The new one produced 57 in three days.

The port worked, and saying so plainly was the useful part. It proved the game was not tied to its engine, and showed what the alternative would cost every day afterwards.

An agent asked to do a port will do the port. It takes a written verdict stage to get one that says don’t.