The Interferometric Field Engine: A Field-Mediated Architecture for Predictive Integration and Two-Regime Recursion in Cortex

The cerebral cortex integrates generative predictions and sensory streams within a 20-watt metabolic budget. Standard hierarchical predictive-coding models locate prediction-error computation in dedicated cortical-circuit machinery. We propose the Interferometric Field Engine (IFE), an architectural pattern that relocates part of this comparison work into the physics of the endogenous ephaptic field. Contributions to the local field carry content-specific signatures that sum by linear superposition, and the resulting interference structure depends on whether same-content contributions match. Through their ordinary nonlinear membrane response, cellular readers recover this as a local dissonance signal that biases plasticity in parallel across the volume, with no central comparator. The architecture’s distinctive commitment is spatial: the interference structure spans the volume at field-propagation speed, which wiring at conduction speed cannot match. The Projected Now is the unified field state at each moment; within it, the Sensed Now is the subset from current sensory drive, the Expected Now is recruited content in conversation with sensation, and the Imagined Now is recruited content running autonomously, where imagery, recall, deliberation, and dreaming live. One comparison physics on these subsets yields two regimes: fast comparison between Sensed and Expected Now, and slow comparison among Imagined Now contributions. Spontaneous thought is then not a separate mode but the slow regime running while the fast regime is quiet. We treat content-specific signatures not as a posited mechanism but as a falsifiable search target, give predictions including temporal-coincidence, closed-loop phase-reset, and propagation-speed tests, and offer a structural critique of Integrated Information Theory’s exclusion postulate