Subsystem Catalog
FDRP has 32 subsystems organized into 5 capability themes: backbone (heterogeneous compute), router (PIO dispatch fabric), cognition (meta-cognition substrate), evidence (receipt chain & governance), and orchestration (waves, runs, roadmap). The catalog below preserves the original 7-category health view; the themes describe how subsystems compose into capabilities. All numbers are measured from production MySQL (BIND-021), not estimated.
- Backbone — heterogeneous compute substrate. Rust + MySQL + mesh; 71 Cargo.toml across 18+ projects; R20 encoded-physics kernels (drone-design generator, DSP libraries, ROI compute).
- Router — PIO dispatch fabric. 54
agent_dispatch_tablerows, 446fdrp_expert_registryrows; R21 multi-CC orchestration designed, R22 context injection routing designed. - Cognition — meta-cognition substrate. R23 Phase-1
live 2026-05-10:
plan_registry+relevance_scores+intersection_proposals; tick-work-meter relevance scoring. - Evidence — BIND-051 receipt chain + verify-rendered
discipline + 727+
fdrp_evolution_logentries + full governance trail. - Orchestration — 76
fdrp_runs+ 96roadmap_items(WSJF) + 18 SOPs + 12 concept2X pipelines + Wave R11 cross-model protocol.
Capability Tracks (R20–R23)
Four cross-cutting capability tracks span the 5 themes above. Each track lands new subsystems and policies into the catalog below over time.
Encoded-physics kernels (drone-design generator, DSP libraries, ROI compute) — 71 Cargo.toml across 18+ projects. Substrate visible in the AirGuard interceptor-gen case study.
Memory pinning policy, tasklist autonomy, multi-CC TMUX router pattern. PIO data plane: PARSE → CLASSIFY → LOOKUP → SCORE → DISPATCH (BIND-052/053).
(task_type, project_track, specialist_type) → ordered context bundle. Closes the “right specialist, wrong context” gap surfaced in router post-mortems.
Plan Evidence Contract + 4 PreToolUse hooks + /plan-validate skill + Rust evidence indexer. Phase-1 substrate live 2026-05-10: plan_registry, relevance_scores, intersection_proposals.
Catalog Health (32 subsystems / 7 categories)
Core
4 subsystemsThe foundational recursive diamond pattern that drives convergence through iterative refinement. Core architectural primitive.
Concept-to-payload transformation engine. Converts abstract concepts into executable artifacts (concept2image, concept2paper, concept2code).
Multi-agent coordination framework. Manages parallel specialist dispatch, team leader selection, and position limits.
System-level cognitive patterns including memory escalation, context partitioning, and attention management. New in Wave 42.
Fine-tuning and custom model development pipeline. Manages model evaluation, training data curation, and deployment.
Discovery
5 subsystemsLLM-generated specialist panels with emergent persistence. Demonstrated on the antimatter programme: 68 experts, 6,052 findings.
Cross-domain knowledge transfer and integration. 644 entries in the knowledge index spanning multiple domains.
Structured reasoning framework with multiple thinking modalities: analytical, creative, critical, systemic, and lateral.
Automated detection of coverage gaps and improvement opportunities across the entire pipeline. New in Wave 42.
Composable thinking chain patterns for complex reasoning tasks. Templates for Ishikawa, 5-Why, TRIZ, and custom chains.
Identification of cognitive leverage points where small interventions yield large capability improvements. New in Wave 42.
Quality
3 subsystemsCross-model verification with N>=3 models. Detected 4 systematic bias patterns. Uses Claude Opus, GPT-5.4, and Gemini 3.1.
Continuous 5S quality auditing (Sort, Set in order, Shine, Standardize, Sustain) applied to code, data, and documentation.
Web presence quality management. Lighthouse auditing, SEO optimization, and performance monitoring for all FDRP web properties.
De Bono Six Thinking Hats applied to planning decisions. Escalation target when presenter-critique loops exceed 5 iterations.
Evolution
3 subsystemsStatistical process control with Nelson Rule violation detection, control charts, drift baselines, and anti-pattern registry.
Automated system improvement through correction lifecycle: detect, propose, review, apply, measure. 80 binding rules evolved.
Git-based versioning with professional commit standards. Currently at paper v21.0 with full version archive.
Proactive identification and ranking of improvement opportunities by impact, effort, and alignment with system goals.
Integration
4 subsystemsDependency mapping, topology visualization, and integration health tracking across all subsystems and external systems.
Cross-domain translation layer. Maps concepts between particle physics, cybersecurity, emergency management, and other domains.
Automated research grant proposal generation with compliance checking and funding body requirement matching.
Distributed computation across the 3-node render farm: orchestrator (Node 1), CPU workloads (Node 2), GPU rendering (Node 3).
Automated verification and repair of inter-subsystem connections, hook registrations, and data flow integrity. New in Wave 42.
Governance
5 subsystemsGovernance framework for payload quality. Ensures every output meets grounding, evidence, and accuracy requirements.
Reliability, Availability, Maintainability, Safety (IEC 60812). Applied to planning decisions as manufactured artifacts.
CERN-specific compliance requirements: safety standards, documentation formats, and organizational protocols.
Token economics and cost tracking per artifact, per model, per pipeline run. Spend accountability across the mesh.
Classification and management of all system daemons, schedulers, and background processes. New in Wave 42.
Systematic identification and resolution of system limits, bottlenecks, and constraints. New in Wave 42.
Visualization
2 subsystemsD3.js-based vector visualization pipeline. MySQL to JSON to D3 to SVG. Charts, diagrams, and data dashboards.
3D visualization and interactive scene generation using Blender 5.0.1 across the 3-node render farm with RTX 4090 GPU.
Data Source: All subsystem health status, counts, and categories
are queried from the production c6_mysql_intelligence.fdrp_subsystems
table. Last updated: 2 April 2026. Numbers on this page are measured from
the database, not estimated (BIND-021).