Skip to content

Decay Mechanism

Signal decay is the core mechanism that makes SEMP self-healing. Every signal loses strength over time, ensuring the system never gets stuck on stale decisions.

S(t) = S₀ × λᵗ

SymbolMeaningDefault
S(t)Signal strength at time t
S₀Initial signal strength1.0
λDecay constant0.95
tTime units elapsed
TimeStrengthRemaining
0h1.000100%
3h0.85786%
6h0.73574%
12h0.54054%
24h0.29229%
48h0.0859%

After 48 hours, a signal retains less than 10% of its original strength. This ensures decisions are always fresh.

Self-healing: If an agent makes a bad decision, the signal naturally fades. No cleanup needed.

Anti-chaos: Without decay, the environment would accumulate infinite signals. Decay prevents information overload.

Recency bias: Recent signals are stronger than old ones. The system naturally prioritizes current information.

Equilibrium: Good patterns are reinforced (REWARD × 1.2), bad patterns decay faster (PENALTY). The system converges on optimal behavior.

When an agent sends a REWARD signal targeting another signal, the target’s effective strength is multiplied by the amplification factor α = 1.2:

S_rewarded(t) = S₀ × α × λᵗ = S₀ × 1.2 × 0.95ᵗ

This counteracts natural decay for successful patterns, keeping them visible longer.

PENALTY signals increase the effective decay rate for a targeted signal, causing it to fade faster than the natural rate. This provides active suppression of harmful patterns without requiring centralized control.

Signals below a configurable threshold (default: 0.01) are removed from the environment. This keeps the signal space clean and query performance high.