Enablement Specifications
Implementation-grade amendments to Prov 1 and Prov 2. Exact sensor thresholds. Sinkhorn convergence bound. Sigmoid permeability mapping. Mahalanobis halt condition. Observable behavioral hesitation. Every parameter has a default. Every bound has a proof. Every safety mechanism has a halt condition.
What it is
The gap between a patent specification and a working system is measured in parameters. Prov 1 describes the minimum gate. This supplement specifies that light quantises at 50/300 lux, sound at 40/65 dBA, and the Sinkhorn-Knopp projector converges to 1.2×10²&sup8; deviation in exactly 20 iterations on 32-bit float. A person skilled in the art can read this filing and build a working CCF implementation without ambiguity.
The sigmoid permeability mapping P(c) = 1/(1+exp(-8(c-0.5))) translates coherence into behavioral permeability with a tension-modulated midpoint shift of ±0.1. The Mahalanobis 3σ halt condition stops the robot when the environment deviates beyond three standard deviations from its learned distribution—when furniture is moved, when lighting changes drastically, when the context becomes unrecognisable.
Claim 29A introduces observable behavioral hesitation: when the reflexive and deliberative processing pathways disagree, the robot visibly slows down. This is not a bug—it is a designed safety signal. Observers can see the robot thinking. The amplitude reduction is proportional to the classification conflict between pathways.
Problems it solves
Moving from theory to buildable system
A patent claim says "sensor quantisation". An engineer asks "what thresholds?" This filing answers: 50/300 lux for light, 40/65 dBA for sound, 0.3/1.5 m for proximity, 0.2/0.8 g for motion, 20/28°C for temperature, 0.1/0.5 Hz for vibration. Six sensor dimensions with exact numeric defaults calibrated for domestic indoor deployment.
Every parameter needs a default
The sigmoid permeability mapping has two parameters: steepness k=8 and midpoint c₀=0.5, with tension modulating the midpoint by ±0.1. These aren't arbitrary—they produce a steep transition that prevents the robot from being "sort of" active in contexts where it should either be engaged or withdrawn.
Every bound needs a proof
The Sinkhorn-Knopp projector converges. But how fast, and to what precision? The Birkhoff contraction factor is τ ≈ 0.403. After 20 iterations on 32-bit float, the maximum row/column sum deviation is bounded by 1.2×10⁻⁸. This is provable and testable on ARM Cortex-M hardware.
Every safety mechanism needs a halt condition
The Mahalanobis distance threshold is 3.0σ. When the environment shifts beyond three standard deviations from the learned distribution signature (mean vector + covariance matrix), the robot halts. Four distinct halt conditions are enumerated. The robot doesn't guess—it stops and reports.
Real-world scenarios
ARM Cortex-M implementation
An embedded engineer builds CCF on an ARM Cortex-M4 with 256KB RAM. The Sinkhorn-Knopp projector uses fixed-size arrays, no heap allocation, and converges in exactly 20 iterations. The sigmoid permeability function compiles to a handful of float operations. Every sensor threshold is specified in the filing. The engineer doesn’t need to tune parameters—the defaults work for domestic indoor deployment out of the box.
Robot halting when furniture is moved
A care home rearranges the common room over the weekend. On Monday morning, the robot’s sensor readings produce a distribution signature that deviates 4.2σ from the learned mean—well beyond the 3.0σ Mahalanobis threshold. The robot halts, reports the environmental change, and enters a re-learning phase. It doesn’t assume the old map still applies. It doesn’t crash. It stops safely and signals that human attention is needed.
Visible hesitation when pathways disagree
A child approaches the robot quickly from behind while it is engaged in a drawing task. The reflexive pathway classifies this as StartledRetreat (sudden proximity change, high instantaneous tension). The deliberative pathway classifies this as QuietlyBeloved (familiar child, high context trust). The classification conflict produces observable hesitation: the robot’s movements slow, its LED shifts to a transitional hue, and it pauses before responding. The child sees the robot “thinking”—a deliberate design choice that makes the robot’s internal state legible.
What the claims cover
Claims 1A, 7A, 14, 29A -- Prov 1 Amendments
Intermediate ceiling constraint (C_eff cannot exceed C_ctx). Quadrant behavioral profiles with hysteresis. Rewritten independent claim with ceiling operation and permeability mapping. Observable behavioral hesitation from pathway conflict.
Claims A, E, I, M-alt, M, O, Z -- Prov 2 Self-Contained Rewrites
Self-contained claim rewrites eliminating parent-claim number dependencies. Hierarchical mixing, ontogenetic identity, privacy-responsive trust, and emergent home context each standalone for prosecution flexibility.
Sections 0033a, 0054a, 0055d, 0058a, 0090a, 0098a -- Enablement Specifications
Sensor quantisation thresholds for 6 dimensions. Sigmoid permeability (k=8, c₀=0.5). Sinkhorn convergence (20 iterations, 1.2×10⁻⁸). Mahalanobis 3σ halt with 4 conditions. Loss function weights for training.
Applications
Licensing enquiries
CCF is released under BSL-1.1 — free for evaluation and non-commercial use. Commercial licensing is available from Flout Labs.
cbyrne@floutlabs.com