Explore the algorithmic foundations of the Aegis Protocol: live Zipf-Mandelbrot token frequency regression, Hawkes point-process diffusion simulation, and Weighted-MSR outlier pruning models.
// 01: TEXT_FREQUENCY_ANALYSIS
Mandelbrot Power-Law Regression
ANALYSIS READY
Language models sample token sequences with smooth power-law decay ($P(r) = P_0(r+\beta)^{-\gamma}$) and low lexical variance. Human writing exhibits natural burstiness and higher vocabulary dispersion.
SAMPLE TEXTS:
Tokens: 0 | Unique: 0
Calculating Zipf-Mandelbrot Power-Law Regression
0.0s elapsed
Determination R²
0.982
Fit threshold > 0.93 = AI
Shannon Entropy
4.82
Bits / token symbol
Type-Token Ratio
0.61
Lexical diversity
Confidence
96.8%
Statistical certainty
Token Rank Frequency Curve
Observed Frequency Mandelbrot Fit-- Zipf Baseline
ANALYSIS RESULT
SYNTHETIC (LIKELY AI-GENERATED)
98.2% SYNTHETIC
// 02: INFORMATION_DIFFUSION
Hawkes Process Diffusion Simulator
MODEL READY
Simulates information diffusion dynamics via self-exciting Hawkes point processes: $\lambda(t) = \mu + \sum_{t_i < t} \alpha e^{-\beta(t - t_i)}$. Computes effective reproduction number $R_0 = \alpha / \beta$ to model network spread and verified corroboration.
SELECT PRESET TOPIC:
Simulating Hawkes Information Diffusion
0.0s elapsed
Reproduction R₀
2.14
Super-critical if > 1.0
Base Intensity (μ)
0.84
Spontaneous arrivals
Excitation (α / β)
1.45 / 0.68
Branching factor
24h Network Reach
34,280
Projected reach
SIMULATION: RED = SPREAD | CYAN = VERIFIED CORROBORATION
Evaluates multi-source verification using Weighted Mean Subsequence Reduction (W-MSR) to prune inconsistent or outlier model assessments before computing consensus.