The Intent Engineering Brief for AI

In 1944, the physicist Erwin Schrödinger asked how a living system maintains order in a universe that trends toward disorder. His answer — negentropy — explains why every AI system you’ve ever used will eventually lie to you.

Large language models generate text one token at a time. Each prediction compounds the uncertainty of every prediction before it. Over long sequences, entropy accumulates. The model drifts. Not because it’s bad. Because that’s what probabilistic systems do. It’s physics.

This brief connects thermodynamic principles to AI governance — why hallucination is structural, why prompt engineering can’t solve it, and what a negentropy engine looks like in practice.