Constructive Interference

The practice

Constructive interference.

physics
Two waves meet in phase; where the crests line up, the signal grows.
practice
A person's voice and a machine's assistance, lined up on purpose, so the writing comes out more itself, not less.

Two waves can meet two ways. Out of phase, they cancel, and the water goes flat. In phase, they add, and the crest doubles. Most writing done with a machine is the first case: the model's cadence and the writer's average into something smoother and emptier than either one alone. Constructive interference is the practice of arranging for the second case—a small set of instruments for keeping a writer's voice, judgment, and understanding intact while they work with a machine, and sharper for it.

The instruments below are built and building. One is finished enough to walk through.

Two waves

Drag to bring them into phase.

out of phase · cancelling

Two signals of the same frequency add by their phase. Aligned, crest reinforces crest and the sum stands twice as tall—constructive. Opposed by half a wavelength, crest meets trough and they erase each other—destructive. Everything between is their exact sum.

The practice puts a person's voice on one wave and the machine's on the other, and works to keep them in phase.

The instruments

Destructive Interference taxonomy · in development

The map of what goes wrong. Eight named classes of structural failure in AI-influenced prose: the emptied intensifier, the false balance, the ending that only restates. Writer-independent—these are failures in anyone's draft, not one person's habits. The point of naming them is discernment. A failure with a name is one you can see coming and cut on sight.

Kiku ่ดใ tool · built

Kiku listens to a finished AI collaboration and pulls out one class of language behavior on command—every place the model slid into caretaking, say, or every place the writer pushed back. Regex catches the obvious hits; a model reads for the ones no pattern could. It is how the practice studies what AI language does across a long working session, with evidence instead of impression.