Skip to content

DEC-0008 — Hybrid Swarm Pattern

Status: Accepted Date: 2026-03-12

Context

The Nexus Method Architect tradition prescribes assessing every project on two dimensions — criticality and complexity — and then prescribing either an Agile Swarm, a Formal Swarm, or a Hybrid Swarm. This decision applies that assessment to the Nexus SDLC framework itself and establishes the baseline pattern that the Methodologist adapts per project.

The initial assessment used Crystal's C0-C3 scale and a Simple/Complex/Highly-Complex scale. DEC-0013 later introduced plain-language profile names (Casual/Commercial/Critical/Vital) that supersede the C0-C3 notation throughout the framework. The hybrid swarm pattern decision itself remains unchanged.

Decision

Framework-Level Assessment

Criticality: Commercial to Critical (project-dependent)

The framework serves projects across the full range — from solo proofs of concept (Casual) to regulated systems (Vital). The framework itself is designed to handle this range. Per-project assessment is performed by the Methodologist at project start.

Complexity: Complex

Multi-agent coordination, persistent artifact trails, human-in-the-loop gates, tool access control, profile-calibrated behavior — this is a distributed system with stateful orchestration. Not Highly Complex (no real-time constraints, no safety-critical hardware integration in the framework itself).

Prescribed Swarm Pattern: Hybrid

The swarm operates as a hybrid of Agile and Formal practices, selected by phase:

  • Agile core: The execution/verification cycle operates with fast feedback, minimal ceremony, and event-driven routing (XP influence).
  • Formal gates: Phase transitions use architecture-gated milestones with explicit quality criteria and Nexus approval (RUP influence).
  • Empirical adaptation: The Methodologist monitors swarm health and reconfigures when trigger events occur (Scrum's empirical process control applied to an agent swarm).
  • Human-centric communication: All Nexus-facing outputs use the Nexus Briefing format, optimized for cognitive efficiency (Crystal influence).

Dominant Methodology by Phase

Phase Dominant Methodology Rationale
Ingestion Crystal Intent elicitation requires cognitive load management and cooperative communication
Decomposition RUP Architecture-centric analysis, risk-first prioritization
Requirements Gate Scrum Transparent inspection of requirements as understood
Architecture Gate RUP Architecture integrity verification before commitment
Plan Gate Scrum Empirical review of proposed plan against risk and value
Execution Cycle XP Engineering discipline, TDD, tight feedback loops
Verification Cycle XP Continuous verification, automated checks
Demo Sign-off Crystal Human-centric summary, trust calibration through running software
Go-Live RUP Milestone quality gate for production readiness

Per-Project Adaptation (via Methodologist)

Profile Adaptation
Casual Minimal agents (Planner, Builder, Verifier). Sketch-weight artifacts. Plan Gate may be informal. Requirements Gate and Architecture Gate may be collapsed or skipped.
Commercial Full agent set minus Designer if no UI. Draft-weight artifacts. All gates active.
Critical Full agent set. Blueprint-weight artifacts. Strict TDD. Fitness functions required. Sentinel active every cycle.
Vital Full agent set. Spec-weight artifacts. Formal Architecture Baseline. Signed-off fitness functions. Nexus explicitly signs off security posture at Demo Sign-off.

Reasoning

Why hybrid over pure Agile: The human gates (Requirements Gate, Architecture Gate, Plan Gate, Demo Sign-off, Go-Live) are inherently formal — they require structured artifacts and explicit approval. Pure Agile would underserve these moments. The Nexus making a Requirements Gate decision needs a structured Briefing, not an informal conversation.

Why hybrid over pure Formal: The execution/verification cycle benefits from speed and minimal ceremony. Formal gates within the iterate loop would add overhead without value. A Builder fixing a failing test does not need a change request.

Why methodology varies by phase: Different phases have different natures. Ingestion is a communication problem (Crystal). Decomposition is a structural analysis problem (RUP). Execution is an engineering discipline problem (XP). Matching the methodology to the phase's nature produces better outcomes than applying one methodology uniformly.

Why per-project adaptation through the Methodologist: The swarm pattern is a baseline, not a fixed configuration. Real projects sit at different points on the criticality and complexity spectra. The Methodologist's job is to read the project and calibrate the swarm — activating the right agents, setting the right artifact weight, and configuring the right gates. A fixed swarm pattern cannot serve both a weekend script and a medical device.

Alternatives Considered

Pure Agile Swarm: Insufficient rigor for the gates. The Nexus needs structured information to make gate decisions. Rejected for under-serving decision quality.

Pure Formal Swarm: Crushing overhead for Casual and Commercial projects. A solo developer building a CLI tool does not need an Architecture Baseline or formal Audit Reports. Rejected for process bloat.

Fixed single pattern (no per-project adaptation): Does not accommodate the range from Casual to Vital. A pattern that works for Casual is reckless for Vital. A pattern that works for Vital is suffocating for Casual. Rejected for inflexibility.

Let the Nexus choose the methodology per phase: The Nexus should not need to understand RUP, XP, Scrum, and Crystal to use the framework. The methodology mapping is embedded in the agent definitions — the Nexus experiences the result (structured gates, fast execution, clear briefings) without needing to know which tradition it draws from. Rejected for imposing unnecessary methodology knowledge on the user.

Consequences

  • The framework serves projects from Casual to Vital without redesign.
  • Each phase uses the methodology best suited to its nature.
  • The Methodologist must encode enough methodology knowledge to make good profile assessments and swarm configurations.
  • The hybrid model is more complex to understand than a pure-style approach — mitigated by the fact that the user does not need to understand the methodology mapping; it is embedded in the agent definitions.
  • C0-C3 and Simple/Complex/Highly-Complex terminology from the initial assessment is superseded by the DEC-0013 profile names (Casual/Commercial/Critical/Vital) throughout the framework.