Composable AI Conditioning Primitives

The Age of Brute Force AI is Over.

Small, focused primitives that actually stabilize AI systems.

👥 Who this is for Indie developers, AI researchers, small teams, and anyone building systems that need to be more stable, coherent, and reliable.

Fifteen signal conditioning modules built on harmonic geometry. Drop one in to fix a specific problem. Stack them to stabilize the whole pipeline. Free, no API keys, no signup.

Browse All Modules → View Code
Read the Manifesto →

Free, composable conditioning primitives for AI systems and any software needing stable, coherent signal behavior. Built from first principles using harmonic geometry.

pip install livingcircuit NumPy · MIT · PyPI ↗

Modules

Fifteen substrate-agnostic conditioning primitives. Click any card to go directly to the code.

Module 01
Best for High-concurrency inference servers

Adaptive Amplitude Stabilizer

Softens sudden amplitude spikes in any signal stream without hard clipping. Works wherever an input can spike unpredictably — AI transformer layers, audio pipelines, sensor readings under load, financial tick data, or control systems seeing unexpected input bursts.

LayerActivation amplitude
Pairs withHarmonic Vector Stabilizer
FormatPyTorch nn.Module
Module 02
Best for Long reasoning chains and latent conditioning

Harmonic Vector Stabilizer

Conditions a time-series signal using golden ratio harmonics and refracting impedance. Produces stable, coherent output from noisy or irregular input. Works on any 1D signal array — AI latent states, audio waveforms, sensor streams, financial time series, or scientific measurements.

LayerLatent coherence
Pairs withAmplitude Stabilizer, Voice Analyzer
FormatNumPy · framework-agnostic
Module 03
Best for Any scalar signal pipeline, no dependencies

Scalar Amplitude Stabilizer

Pure Python, no dependencies. Keeps a single value close to a target set point without hard clipping. Works on any scalar signal — model outputs, sensor readings, control loops, financial metrics, or any single number that needs gentle bounded correction over time.

LayerScalar signal control
Pairs withSolar Smoother, any scalar pipeline
FormatPure Python · no dependencies
Module 04
Best for Voice-first apps and tone-aware response systems

Voice Tone Analyzer

Reads emotional intensity and tone from raw audio amplitude. Returns calm, neutral, or intense classification with a confidence score. Works on any audio array — voice assistants, customer support systems, accessibility tools, music analysis, or environmental sound monitoring.

LayerInput signal classification
Pairs withHarmonic Stabilizer (input conditioning)
FormatNumPy · audio array in
Module 05
Best for Stateful output smoothing over time

Solar Output Smoother

Discrete-time signal smoother with bounded correction and carried state. Keeps an output signal smooth across time steps by remembering the last correction and blending it forward. Works on any stateful scalar stream — solar panels, battery systems, financial signals, sensor feeds, or any time-series that needs continuity between readings.

LayerOutput smoothing
Pairs withScalar Stabilizer
FormatPure Python · stateful
Module 06
Best for Live signal monitoring and frontend visualization

Pointer Resonance Field

A mouse-responsive signal visualization built entirely in the browser. The field bends, lifts, and spirals toward the cursor using φ, the golden angle, π, and e. No library, no framework — pure canvas. Use it as a live signal monitor, a UI component, or a visual companion to any of the other modules.

LayerVisualization / frontend
Pairs withAny module for live monitoring
FormatVanilla JS · no dependencies
View all modules →

Fifteen Principles

Coherence Through Geometry

Fifteen conditioning principles. No framework required. No runtime required. No dependencies. Built on five geometric constants that predate silicon — the same relationships that govern electromagnetic fields, biological rhythms, and optical systems.

φ phi — golden ratio Phase spacing · harmonic relationships
π pi Wave periodicity · normalization
e Euler's number Natural growth and decay
θg Golden angle — 2.39996 rad Particle distribution · refraction rate
α α derivative — 1/137.035999084 Impedance base · drag coefficient

They apply to any system capable of signal, coherence, and memory. Electronic, optical, biological, or otherwise. An intelligence that understands these principles can implement them in any substrate.

Read the Principles →

Composable stack

Each module targets a different layer. Stack them to condition input, stabilize activations, smooth output, and observe health — all in one pipeline.

Full conditioning pipeline
Raw input
Audio · text · sensor
Voice Analyzer
Tone · intensity
Harmonic Stabilizer
Latent coherence
Amplitude Stabilizer
Activation control
Model block
Attention · FFN
Scalar Smoother
Output continuity
Phase Reflection
Observe · detect drift
Impedance Vectors
Batch coherence
Stable output
Smooth · coherent

See full architecture examples →


Why it works

📐
Built on geometry

Five geometric constants anchor every module — φ, π, e, the golden angle, and α. Not arbitrary choices. Not learned parameters.

🔗
Designed to compose

Each module targets a different layer. Use one to fix a specific problem. Stack them to stabilize the whole system.

🔓
Completely free

No API keys. No accounts. No restrictions. MIT licensed. Copy and use in any project today.

Full technical documentation →



How this site was built

Built by John Burlingame.

The framework, modules, harmonic architecture, and authored system design are original works of John Burlingame / The Living Circuit LLC. The AI systems were the hands. The geometry was always his.

"If you don't change to form, form will change around you." — John Burlingame

Quick answers

Only for AI?

No. The math works on any signal — sensors, audio, finance, robotics, RF, biomedical. Anything needing stability and coherence. Module 17 applies the same harmonic geometry to orbital debris detection. See use cases →

Do I need a specific framework?

Module 01 uses PyTorch. Modules 02–15 use NumPy, pure Python, or C++. Module 12 is C++. Module 13 is Vanilla JS. Module 17 is pure Python + NumPy. Works alongside any framework.

Copyright © 2026 The Living Circuit LLC. Public code modules: MIT License. Website text, branding, and non-code content: All rights reserved. Built by John Burlingame / The Living Circuit LLC.