Cybernetics

Cybernetics was coined by Norbert Wiener in 1948 to describe the study of control and communication in animals and machines. The word derives from the Greek kybernetes (steersman). Wiener noticed that effective regulation — whether in a thermostat, a nervous system, or a factory floor — shares the same abstract structure: a goal, a sensor that measures the current state, and a feedback loop that closes the gap.

• Feedback — the mechanism by which a system uses information about its own output to correct its future behaviour. Negative feedback (error-reducing) produces stability; positive feedback (error-amplifying) produces growth or collapse.
• Variety — W. Ross Ashby's term for the number of distinguishable states a system can be in. Ashby's Law of Requisite Variety states that only variety can absorb variety: a regulator must be at least as complex as the disturbances it is meant to handle.
• Homeostasis — the tendency of a viable system to maintain essential variables within tolerable bounds in the face of environmental perturbation.
• Ultrastability — Ashby's term for a system with two feedback loops rather than one. The first regulates within a fixed configuration. The second changes the configuration itself, and acts only when the first fails to hold an essential variable within its bounds. Homeostasis maintains; ultrastability reorganises when maintenance fails. Ashby's homeostat demonstrated the arrangement with a fast continuous loop that adjusted its outputs and a slow discrete loop that stepped to new parameters once limits were breached. The practical import is that a regulator built only for the disturbances its designer foresaw is defeated by any other, and that adaptation requires the failure itself to act as a signal.
• Regulatory function — one of the functions a regulator must discharge in order to regulate at all. Wiener's decomposition names three: a goal against which the state is compared, a sensor by which the state is known, and a path by which action is taken. Ashby's law adds a fourth requirement, that the regulator command sufficient variety, and the good regulator theorem a fifth, that it hold a model of what it regulates. A regulator lacking any of them fails in a specific way: with no goal there is nothing to compare against, with no sensor no error can be detected, with no effector no correction can be made, with insufficient variety the disturbance is not absorbed, and with no model the response cannot be chosen. The phrase is not one author's term of art; it names the composition of these requirements, and it is used in this project to classify the documents that supply each.
• The good regulator theorem — Conant and Ashby proved that a regulator which is both maximally successful and maximally simple must be isomorphic to the system it regulates: every good regulator of a system must be a model of that system. The theorem is usually cited for the first half, that effective regulation requires a model. The simplicity clause matters equally, since it excludes modelling anything the regulator does not regulate.
• Black box — the practice of characterising a system purely by its inputs and outputs, without needing to know its internal structure. Useful when the internals are inaccessible or irrelevant to the regulatory task.

Stafford Beer applied Wiener's framework to human organisations in the 1950s–70s. His central question was: why do large organisations so frequently fail to regulate themselves? His answer: they are not structured to produce the variety necessary to absorb the variety of their environments. A command-and-control hierarchy collapses information before it reaches decision-makers and delays action until it is too late.

Beer's response was the Viable System Model — a recursive, five-level architecture whose key property is that every level is itself a viable system. The model does not describe a hierarchy of authority; it describes a separation of time horizons and concerns.

ORE Studio uses Beer's VSM as a lens, not a blueprint. The five systems map onto the roles that LLM agents, sprint planners, and the product identity play in the development process. The mapping is intentionally loose — the point is not to implement VSM faithfully but to borrow its language for separating concerns:

• Concerns that change within hours belong to S1 Agent (task execution).
• Concerns that change within days belong to S2 Orchestrator (story orchestration).
• Concerns that change within weeks belong to S3 Sprint Planner (sprint management).
• Concerns that change within months belong to S4 Version Planner (version planning).
• Concerns that are durable belong to S5 Identity Steward (product identity).

For where this mapping aligns with and departs from canonical VSM, see the Connection to and divergence from the VSM section of Cybernetic Levels.