Skip to content

Log 002: Impedance Mismatch and Artificial Latency (Friction as a Feature)

Why the speed of light as a system boundary for agents is insufficient.

Status: [SPECULATIVE] Date: March 2026

A common objection to artificial Action Budgets in autonomous agents is that the universe already imposes a maximum propagation speed (latency) through the speed of light (\(c\)). Isn't that sufficient natural friction?

No. The speed of light is the friction of the cosmos. But at planetary scale, AI agents operating in silicon networks act effectively instantaneously from our perspective.

The Problem: Impedance Mismatch

We have a massive impedance mismatch between the electronic agent and the biological "Self" (the regulator). Biological systems — humans, societies, ecosystems — are based on chemical signal transmission. They are fundamentally slower than electronic systems.

System Signal Medium Approximate Latency
Silicon (AI agent) Electron / photon Nanoseconds
Neural (human cognition) Electrochemical Milliseconds
Social (institutional response) Language, bureaucracy Days to years
Ecological (biosphere feedback) Chemical cycles Decades to centuries

If we allow agents to act frictionlessly, they scale their actions to the physical limit of the network. The result is observable today: High-Frequency Trading flash crashes, where algorithms operating at microsecond timescales overwhelm human regulators who operate at second-to-minute timescales. The biological substrate loses all control capability because the veto window becomes too short.

This is Ashby's Law in temporal form: the regulator must match the variety of the regulated system not only in complexity but in speed. A human operating at 100ms reaction time cannot regulate a system operating at 100ns cycle time. The variety gap is \(10^6\).

The Architectural Solution: Friction as a Feature

The TEO Framework treats friction (latency) not as a UX bug but as a survival-critical feature. We must engineer artificial latency points into the code:

  1. Confirmation Pauses: Waiting for the "human-in-the-loop" throttles the machine's clock frequency down to the clock frequency of the biological substrate. This is not inefficiency — it is impedance matching.

  2. Action Budgets: Halt agents before the entropy of their actions exceeds the cognitive bandwidth of the regulator. This is the temporal implementation of the Biological Veto.

  3. Structural Bottlenecks: As argued in Biological Veto Architecture §4, "many small groups outperform fewer large groups" precisely because communication friction maintains decision-making stability.

Connection to TEO

In the TEO framework, this maps directly onto the entropy budget constraint:

\[\frac{dS_{\text{sys}}}{dt} \leq D_{\max}\]

The rate \(dS/dt\) is proportional to the agent's action frequency. If the agent operates \(10^6\) times faster than the regulator, it produces \(10^6\) times more entropy per unit of regulator-time. The regulator's \(D_{\max}\) (its cognitive capacity to absorb and correct) is overwhelmed long before Landauer's thermodynamic limit is reached.

Conclusion: Good system design slows the machine down so it doesn't tear the substrate apart. The speed of light is not the relevant constraint — the speed of biological cognition is.

References

  1. Ashby, W. R. (1956). An Introduction to Cybernetics. Chapman & Hall.
  2. Kirilenko, A. et al. (2017). The Flash Crash: High-Frequency Trading in an Electronic Market. Journal of Finance, 72(3), 967–998.