The Living Fingertip

click/drag = touch the skin
§Z₂ cyan: flash at press/release
§SO₂ purple: glow while held
Mexican hat: σ ratio = 2
CRR (A. Sabine, pending peer review) · temporalgrammar.ai · Active Inference Institute

The Fingertip as Active Inference

This fingertip is built from CRR (Coherence-Rupture-Regeneration). The fingerprint pattern, the receptor responses, and the heatmap dynamics all emerge from three equations and one parameter.

The Skin as Markov Blanket

The skin IS the body's Markov blanket: the statistical boundary separating internal from external states. The epidermis (sensory surface) faces the world. The dermis (internal states) contains the generative model machinery. Four mechanoreceptor types sit at this boundary.

F = DKL[q(s) || p(s|o)] - ln p(o)The fingertip minimises free energy through active touch (exploration)

Z₂ Sensory: Rapidly Adapting

Meissner corpuscles (RA1) and Pacinian corpuscles (RA2) detect change only. They fire at onset and offset: a bistable Z₂ operation. The cyan flash you see when you press down IS the Z₂ rupture.

Z₂ channel (likelihood precision):
ΩZ₂ = 1/π ≈ 0.318 · C* = πOnset/offset binary. Rapidly adapting. Detects CHANGE.

Meissner corpuscles account for ~40% of hand innervation. The Pacinian corpuscle's concentric lamellae are literally S¹: the SO(2) precision filter made physical in connective tissue.

SO(2) Prior: Slowly Adapting

Merkel cells (SA1) and Ruffini endings (SA2) provide sustained monitoring. They fire continuously while pressure is applied: a continuous SO(2) accumulation. The purple glow that builds while you hold IS the SO(2) coherence.

SO(2) channel (transition precision):
ΩSO(2) = 1/2π ≈ 0.159 · C* = 2πSustained firing. Slowly adapting. Sets PRECISION.

Merkel cells have an inhibitory surround that sharpens spatial resolution. This is the SO(2) lateral inhibition that creates the Mexican hat.

The Mexican Hat

At the cuneate nucleus in the brainstem, feedforward lateral inhibition creates centre-surround receptive fields. Excitatory centre (Z₂, σe) surrounded by inhibitory flanks (SO(2), σi).

σsurround / σcentre = ΩZ₂ / ΩSO(2) = 2The same ratio as in vision and hearing. Topological invariant.

The Fingerprint is CRR-Solved

The fingerprint ridges are a Turing pattern: a standing wave at criticality emerging from the Z₂/SO(2) Mexican hat interaction. CRR derives the ridge spacing with 2.5% accuracy:

λ = 2π · σe · √(2·ln(2) / 3) = 4.27 × σe
σe ≈ 0.12mm (Merkel RF) → λ ≈ 0.51mmKnown ridge spacing: ~0.50mm. Zero free parameters beyond anatomy.

Each ridge IS the beauty function B(C) = exp(C/Ω)·(C*−C) evaluated spatially. The ridge peak sits at C* − Ω (89.9% of capacity): maximum poise before rupture. The valley between ridges is a rupture scar where coherence reset to zero.

Pattern type depends on finger pad curvature: high curvature → whorl (SO(2) dominant), moderate → loop, low → arch (Z₂ dominant). Geometry sets boundary conditions; CRR generates the pattern.

Two-Point Discrimination

Fingertip: 2mm (Ω ≈ 0.32, high precision)
Back: 40mm (Ω ≈ 6.4, low precision)22× dynamic range (27 dB). Spatial Ω map across the body.

References

Parr, T., Pezzulo, G. & Friston, K. (2022). Active Inference. MIT Press.

Abraira, V. E. & Ginty, D. D. (2013). The sensory neurons of touch. Neuron, 79, 618-639.

Johansson, R. S. & Vallbo, Å. B. (1983). Tactile sensory coding in the glabrous skin. TINS, 6, 27-32.

Kucken, M. & Newell, A. C. (2005). Fingerprint formation. J. Theor. Biol., 235, 71-83.

Sabine, A. (2026). Phase-gating across precision channels. AGI-26. temporalgrammar.ai

CRR pending peer review · Active Inference Institute · 2026