Quantm ships with a hard separation between presentation and engine. The light and dark themes are entirely cosmetic — they re-paint the surface without touching a single line of the synthesis pipeline. A derivation produced in dark mode and a derivation produced in light mode against the same input are byte-identical, because the palette is downstream of the engine and the engine is upstream of every visual choice.
This separation matters more than it sounds. Most products entangle interface state with engine behaviour to a surprising degree — a layout change quietly alters which content is included, a theme change quietly affects which renderer is used, a viewport change quietly downgrades which derivation is run. Quantm refuses that entanglement by construction. The presentation layer reads from the engine; it never writes to it.
The practical consequence is that a user who works half the day in light mode and the other half in dark mode receives identical engine output across both. The interface adapts to the environment; the deterministic accuracy does not negotiate. That is the only sensible relationship between a presentation layer and a logic core, and it is the relationship Quantm is built to enforce.