Skip to content

Latest commit

 

History

History
454 lines (302 loc) · 19.6 KB

File metadata and controls

454 lines (302 loc) · 19.6 KB

VSM Glossary for PACT

Purpose: Define Viable System Model terminology in the context of the PACT framework.

Usage: Reference when reading VSM-enhanced PACT documentation or when VSM terms appear in protocols, agents, or commands.

Created: 2026-01-10


Core Systems

S1 — Operations

VSM Definition: The primary activities that produce value. The "muscles and organs" of the organization—the parts that do the actual work.

In PACT: The specialist agents that execute development tasks:

  • pact-preparer — Research and requirements gathering
  • pact-architect — System design and component planning
  • pact-backend-coder — Server-side implementation
  • pact-frontend-coder — Client-side implementation
  • pact-database-engineer — Data layer implementation
  • pact-test-engineer — Testing and quality assurance
  • pact-n8n — Workflow automation

Key Point: S1 units need autonomy to respond to their local environment while remaining part of the coherent whole.


S2 — Coordination

VSM Definition: The coordination function that resolves conflicts between S1 units, establishes shared language, and dampens oscillations. Acts as a communication conduit between S1 and S3.

In PACT: Coordination mechanisms including:

  • pact-s2-coordination.md — Shared coordination protocols
  • Decision logs — Standardized documentation
  • Pre-parallel conflict detection — Identifying shared resources before parallel execution
  • Resolution authority — Who decides when agents disagree

Key Point: S2 enables S1 autonomy by handling inter-unit coordination so each unit doesn't need to negotiate directly with every other unit.


S3 — Control (Operations Management)

VSM Definition: Manages the ongoing business of S1-S2-S3. Allocates resources, sets performance expectations, optimizes operational efficiency. Focuses on "inside and now."

In PACT: The Orchestrator's operational control function:

  • Task execution management
  • Agent coordination and sequencing
  • Resource allocation (which agents, how many)
  • Progress tracking
  • Blocker resolution (via imPACT)

Key Point: S3 asks "Are we doing things right?" — focused on efficient execution of the current plan.


S3* — Audit Channel

VSM Definition: A sporadic audit channel that allows S3 to directly monitor S1, bypassing S2's filters. Catches what routine coordination might miss. The "sporadic" aspect refers to unpredictability (preventing gaming), not real-time concurrent monitoring.

In PACT: Independent observation mechanisms:

  • Risk-tiered testing — Higher scrutiny for security-sensitive, novel, or complex code
  • Mandatory uncertainty coverage — Coder-flagged concerns MUST be tested
  • Automated checks — Lint, type checking, security scans (static tooling)
  • Direct quality signals — Test engineer reports GREEN/YELLOW/RED to orchestrator

Architectural Note: Due to Claude Code's agent isolation (agents run in separate 200K-token context windows), S3* cannot be realized as parallel real-time monitoring. Instead, PACT implements S3* through:

  1. Strengthened sequential TEST phase with risk-based rigor
  2. Mandatory coverage of uncertainty areas flagged by coders
  3. Static analysis tools that can run independently

Key Point: S3* is about independent observation and ground truth, not timing. While not achieving real-time monitoring, this sequential approach provides independent ground-truth observation. The test engineer's independence from coders (sequential execution) and mandatory uncertainty coverage approximate the audit function that S3* represents in classical VSM. In PACT's sequential model, the VSM concept of "unpredictability" (which prevents gaming the audit) is approximated through mandatory uncertainty coverage: coders cannot predict which of their flagged uncertainties the test engineer will scrutinize most heavily, preserving the audit's independence.


S4 — Intelligence (Development)

VSM Definition: Looks at the outside world, scans for threats and opportunities, creates plans for long-term viability. Focuses on "outside and future."

In PACT: The intelligence and planning function:

  • plan-mode — Strategic planning before implementation
  • pact-preparer — Environment scanning, requirements research
  • Adaptation checks — "Should we change course?"
  • Risk assessment — "What could go wrong?"

Key Point: S4 asks "Are we doing the right things?" — focused on whether the plan still makes sense given external reality.


S5 — Policy (Identity)

VSM Definition: Embodies the identity of the organization—values, norms, ethics, culture. The highest decision-making authority. Provides ground rules and enforces them. Balances the tension between S3 and S4.

In PACT: The governance layer:

  • User as ultimate authority
  • CLAUDE.md principles (formalized)
  • Non-negotiables (SACROSANCT rules)
  • Policy checkpoints
  • Arbiter when S3/S4 conflict

Key Point: S5 doesn't manage operations—it defines what the system IS and what it will NOT do, regardless of operational pressure.


Key Concepts

Algedonic Signal

VSM Definition: An emergency signal (from Greek algos = pain, hedos = pleasure) that bypasses the normal management hierarchy. Like pain signals going straight to the brain, these indicate something requiring immediate attention.

In PACT: Critical signals that bypass normal orchestration:

  • HALT signals: Security vulnerabilities, data exposure, ethical violations → immediate stop
  • ALERT signals: Repeated failures, fundamental misunderstandings → immediate attention

Contrast with imPACT: imPACT is S3's exception handling (triage within normal flow). Algedonic signals bypass S3 entirely, going direct to S5 (user).

Key Point: Not every problem is algedonic. Reserve for viability-threatening situations where normal channels are too slow or might filter the signal.


Autonomy

VSM Definition: The capacity of an S1 unit to adapt its behavior based on local conditions without requiring permission from higher systems for every action.

In PACT: Specialist agents' authority to:

  • Adjust implementation approach based on discoveries
  • Request additional context from other specialists
  • Recommend scope changes
  • Invoke nested PACT cycles for complex sub-tasks

Bounded by: Escalation requirements (contradicts architecture, exceeds scope, security implications)

Key Point: Autonomy isn't independence—it's the freedom to adapt within defined boundaries while remaining part of the coherent whole.


Cohesion

VSM Definition: The property of the whole system acting as a unified entity despite autonomous parts. Achieved through S2-S5 functions, not through eliminating S1 autonomy.

In PACT: Maintained through:

  • S5 policy constraints (non-negotiables)
  • S2 coordination protocols (shared language, conflict resolution)
  • S3 operational management (sequencing, resource allocation)
  • Common goals and identity

Key Point: The goal is autonomy WITH cohesion, not autonomy OR cohesion.


Cybernetic Isomorphism

See: Recursion


Homeostasis

VSM Definition: The dynamic equilibrium maintained by a viable system. Not static balance, but continuous adjustment to maintain stability.

In PACT: Maintained through:

  • S3/S4 tension (operations vs adaptation) balanced by S5
  • Feedback loops (test results, blocker reports, audit signals)
  • Variety management (matching response capacity to task complexity)

Key Point: Homeostasis isn't the absence of change—it's stability through continuous adaptation.


Recursion (Cybernetic Isomorphism)

VSM Definition: Viable systems contain viable systems. Each S1 unit can be modeled using the identical VSM structure. A department is a viable system within a company, which is a viable system within an industry.

In PACT: Nested PACT cycles:

  • A complex feature can contain sub-features, each with own P→A→C→T cycle
  • A specialist can invoke nested PACT for complex sub-tasks
  • The same structure applies at every level

Nesting Limit: 1 level in PACT (to prevent infinite recursion)

Key Point: Recursion means the same principles and structure apply at every scale—you don't need different management theories for different levels.


Requisite Variety (Ashby's Law)

VSM Definition: "Only variety can absorb variety." A controller must have at least as much variety (range of possible responses) as the system it's trying to control. If the environment has 100 possible states, the controller needs at least 100 possible responses.

In PACT: Variety management:

  • Task variety: Novelty, scope, uncertainty, risk
  • Response capacity: Available specialists, parallelization, tools, precedent
  • Attenuators: Reduce incoming complexity (standards, templates, decomposition)
  • Amplifiers: Increase response capacity (more agents, nested cycles, research)

Key Point: You can't control a complex situation with a simple response. Either simplify the situation (attenuate) or increase your response capacity (amplify).


Variety

VSM Definition: The number of possible states a system can be in. A measure of complexity. High variety = many possible states = more complex.

In PACT: Task complexity dimensions:

  • Novelty: Routine → Familiar → Novel → Unprecedented
  • Scope: Single concern → Cross-cutting
  • Uncertainty: Clear → Ambiguous → Unknown
  • Risk: Low → Critical

Key Point: Variety isn't good or bad—it's a measure that helps you match your response appropriately.


Variety Amplifier

VSM Definition: Something that increases the variety (response capacity) of a controller. Allows the system to handle more complex situations.

In PACT Examples:

  • Invoking additional specialists
  • Enabling parallel execution
  • Running nested PACT cycles
  • Research phase to build understanding
  • AI-assisted tools

Variety Attenuator

VSM Definition: Something that reduces the variety (complexity) coming into a system. Simplifies what the controller has to deal with.

In PACT Examples:

  • Applying existing patterns/templates
  • Decomposing into smaller sub-tasks
  • Constraining scope to well-understood territory
  • Using standards to reduce decision space
  • Bounded contexts

Viability

VSM Definition: The capacity of a system to maintain a separate existence—to survive and thrive in a changing environment. Not just current survival, but ongoing adaptability.

In PACT: A development workflow that can:

  • Complete current tasks successfully
  • Adapt when requirements change
  • Handle unexpected complexity
  • Recover from failures
  • Maintain quality under pressure
  • Evolve over time

Key Point: Viability isn't efficiency—a highly efficient but brittle system isn't viable. Viability includes resilience and adaptability.


Operational Terms

These terms are specific to PACT's implementation of VSM concepts.

Environment Model

Definition: An explicit documentation of the assumptions, constraints, and context that inform S4's assessment of plan validity. Created during PREPARE phase, referenced during S4 checkpoints.

Location: docs/preparation/environment-model-{feature}.md

Contents:

  • Tech stack assumptions (language, framework, dependencies)
  • External dependencies (APIs, services, data sources)
  • Constraints (performance, security, time, resources)
  • Unknowns (acknowledged gaps, questions needing answers)
  • Invalidation triggers (what would force approach changes)

When Required:

  • Variety 11+: Required (high complexity demands explicit tracking)
  • Variety 7-10: Recommended (document key assumptions)
  • Variety 4-6: Optional (implicit model often sufficient)

Key Point: The Environment Model makes implicit assumptions explicit. S4 checkpoints compare current reality against this baseline to detect divergence.


Decision Log

Definition: Standardized documentation produced during the CODE phase that captures implementation decisions, rationale, and context for subsequent phases.

Location: docs/decision-logs/{feature}-{domain}.md

Contents:

  • What was implemented and why
  • Key decisions and trade-offs
  • Assumptions made
  • Known limitations
  • Areas of uncertainty (where bugs might hide)

Key Point: Decision logs explain the "why" not the "what"—code shows what was done, the log explains the reasoning.


Orchestration Decision Log

Definition: S3-level audit trail maintained by the orchestrator during /PACT:orchestrate runs. Captures orchestration decisions (variety assessment, agent selection, phase outcomes, S4 checkpoints) for retrospective analysis and pattern recognition.

Location: docs/decision-logs/orchestration-{feature}.md

Distinction from Decision Log:

  • Decision Log: Created by agents during CODE phase, documenting implementation decisions
  • Orchestration Decision Log: Created by orchestrator, documenting orchestration decisions (what agents to invoke, how to coordinate, when to adapt)

Format Tiers:

  • Variety 7-9: Lightweight (key decisions only)
  • Variety 10+: Full format (complete audit trail with S3/S4 tensions, algedonic signals, retrospective)
  • Variety 4-6: None (comPACT territory—too simple for orchestration log)

Key Point: The orchestration log provides meta-level visibility into how the orchestration was conducted, separate from what was implemented.


META-BLOCK

Definition: An ALERT-level algedonic signal category triggered when 3+ consecutive imPACT cycles fail to resolve the same blocker.

Triggers: Same blocker recurring, systemic issue detected, unable to make progress despite multiple attempts.

Response: User attention required—may indicate fundamental misunderstanding or need to restart from an earlier phase.

Key Point: META-BLOCK is proto-algedonic—it starts as operational (imPACT) but escalates to viability concern (ALERT) through repetition.


Override Protocol

Definition: The procedure for continuing work after a HALT signal when the user explicitly chooses to proceed despite identified risks.

Requirements:

  1. Acknowledge the specific risk (not just "I understand")
  2. Explain why proceeding is acceptable
  3. Accept responsibility for consequences

Documentation: Logged in session notes and decision log with "⚠️ Overrode {category} HALT: {justification}"

Key Point: Overrides don't carry forward—if the risk materializes later, a new HALT is required.


S3/S4 Tension

Definition: The inherent conflict between S3 (operational control: "execute now") and S4 (strategic intelligence: "are we doing the right thing?"). This tension is natural and healthy, but unrecognized tension leads to poor decisions.

Common Manifestations:

  • Schedule vs Quality (skip phases vs thorough work)
  • Execute vs Investigate (code now vs understand first)
  • Commit vs Adapt (stay course vs change approach)
  • Efficiency vs Safety (speed vs coordination overhead)

In PACT: When tension is detected:

  1. Name it explicitly
  2. Articulate trade-offs for each path
  3. Resolve based on project values or escalate to user (S5)

Key Point: S5 is the arbiter of S3/S4 tension when values alone don't resolve it. The user decides which trade-off is acceptable.


Research Spike

Definition: A time-boxed exploration activity recommended for extreme variety tasks (score 15-16) to reduce uncertainty before committing to implementation.

Purpose: Reduce task variety by building understanding, testing assumptions, and mapping unknowns.

Outcome: After the spike, reassess—the task should now score lower. If still 15+, decompose further or reconsider feasibility.

Key Point: A spike is not implementation—it's reconnaissance. The goal is reducing variety, not producing code.


Temporal Horizon

Definition: The characteristic time scale at which each VSM system operates. Different systems naturally focus on different planning and decision horizons.

In PACT:

System Horizon Focus
S1 Minutes Current subtask (agent implementation)
S2 Parallel dispatch Coordination across parallel specialists (boundary/convention enforcement during concurrent dispatch)
S3 Hours Current task/phase (orchestrator coordination)
S4 Days Current milestone/sprint (planning, adaptation)
S5 Persistent Project identity (values, non-negotiables)

Key Point: When you find yourself in S3 mode asking S4-horizon questions ("will this scale next quarter?"), you're experiencing mode/horizon misalignment. Recognize it and either transition modes or note the question for later.


Variety Score

Definition: A 4-16 numeric assessment of task complexity used to select the appropriate workflow ceremony level.

Dimensions (each scored 1-4):

Dimension 1 (Low) 4 (Extreme)
Novelty Routine Unprecedented
Scope Single concern Cross-cutting
Uncertainty Clear Unknown
Risk Low impact Critical

Thresholds:

Score Level Workflow
4-6 Low comPACT
7-10 Medium orchestrate
11-14 High plan-mode → orchestrate
15-16 Extreme Research spike → Reassess

Key Point: Variety score guides ceremony level—higher variety needs more preparation and coordination.


Quick Reference Table

Term One-Line Definition PACT Equivalent
S1 Primary operations Specialist agents
S2 Coordination between S1 units Protocols, conflict resolution
S3 Operational control (inside-now) Orchestrator execution mode
S3* Audit channel bypassing S2 Independent quality observation
S4 Intelligence (outside-future) plan-mode, adaptation checks
S5 Policy/identity/values User + CLAUDE.md principles
Algedonic Emergency bypass signal HALT/ALERT to user
Autonomy Local adaptation authority Agent autonomy charter
Cohesion System unity despite autonomy Shared protocols, policy
Homeostasis Dynamic equilibrium S3/S4 balance via S5
S3/S4 Tension Operational vs strategic conflict Name, trade-off, resolve/escalate
Recursion Viable systems within viable systems Nested PACT cycles
Requisite Variety Controller needs matching complexity Variety budget assessment
Variety Measure of complexity/possible states Task complexity dimensions
Viability Capacity for ongoing existence Adaptive, resilient workflow
Environment Model Explicit assumptions and constraints docs/preparation/environment-model-*
Decision Log Implementation documentation docs/decision-logs/
Orchestration Decision Log S3-level orchestration audit trail docs/decision-logs/orchestration-*
META-BLOCK 3+ imPACT cycles → ALERT Escalation to user
Override Protocol HALT continuation procedure Justified risk acceptance
Research Spike Extreme variety exploration Pre-implementation recon
Temporal Horizon Time scale for each VSM system S1=min, S2=concurrent, S3=hrs, S4=days, S5=persistent
Variety Score 4-16 complexity assessment Workflow ceremony selector

Further Reading

  • Beer, Stafford. Diagnosing the System for Organizations (1985) — Most accessible introduction
  • Beer, Stafford. Brain of the Firm (1972) — Original VSM formulation
  • Beer, Stafford. The Heart of Enterprise (1979) — Detailed theoretical treatment
  • Metaphorum VSM Resources