System Architecture

Neurotrophic Operating System — from signal to resolution

v1.0 2026-03-26 PUBLIC

A deterministic, identity-free substrate that processes signals through a governed collapse pipeline. Every input either resolves to a unique fixed point or is vetoed. Nothing persists uncommitted. Nothing executes ungoverned.

The system operates as a Neurotrophic OS — a substrate that maintains its own vitality, adapts its own structure, repairs its own faults, learns from its own history, and governs its own learning. External systems connect through certified adaptors.

System Layers

Every signal travels one path, top to bottom. No branches. No ambiguity.

External World — real-time data, simulation inputs, human decisions Connector ingress boundary and Atom sovereign dispatch kernel Collapse Pipeline — five sequential irreversible transformations Governance Filtration and Topology seed graph

Collapse Pipeline

Five sequential, irreversible transformations. Each signal enters the pipeline once and exits as either a resolved output or a veto. No branching. No partial pass.

αAdmissibility
νValidation
λRouting
ρReaction
εExtinction
α Is this signal admissible? Binary gate, no partial pass.
ν Collapse possibility to well-formed element.
λ Collapse to exactly one reaction path. No branching.
ρ Deterministic reaction. Same input → same output.
ε Extinguish to purity (ω₀). Nothing persists uncommitted.

C² = C — Collapse is a projection. Applying it twice changes nothing.
C¹⁻ ∄ — Collapse is irreversible.
Six veto classes can reject at any stage. Vetoed signals are structurally impossible, not filtered.

Governance Filtration

Constitutional layers. Each inherits constraints from below. Doctrine loaded at boot, immutable at runtime. No runtime self-modification of governance rules.

G₀ ⊆ G₁ ⊆ G₂ ⊆ G₃ ⊆ G₄

Constraint Surface

  • FROZEN — defines what is admissible
  • Cannot be modified at runtime
  • Loaded from doctrine at boot

Adaptation Surface

  • MUTABLE within bounds
  • Parameters the neurotrophic layer can modify
  • Bounded by the constraint surface

Neurotrophic Layer — Governed Adaptation

The neurotrophic layer sits above the fast path and observes its output. It never interrupts the fast path. It learns from it, adapts parameters within bounds, and feeds changes back — but the fast path continues running unchanged through every adaptation cycle.

Neurotrophic Layer — fast path and slow path with Phases A through E

 Phase A — Governed Homeostasis

Observes fast-path thermodynamic output (rate, latency, error, throughput). Evaluates against doctrine-defined set points. Adjusts operational parameters within constitutional bounds. The slow path watches, measures, reweights — the fast path runs unchanged.

 Phase B — Structural Plasticity

Grows and prunes topology nodes. Adds new seeds, strengthens connections, removes underperforming paths. All growth bounded by the frozen constraint surface. This is how the system’s structure evolves without violating its constitution.

 Phase C — Cross-Domain Fault Repair

Astrocyte-model mediator. Detects faults across domain boundaries. Propagates bridge signals. Repairs damaged circuits before learning proceeds. Learning on damaged circuits produces drift — Phase C closes that gap.

 Phase D — Temporal Learning

Learns from history. Applies reward mechanisms. Balances beneficial outcomes against hostile ones. Accumulates evidence until prediction confidence reaches threshold. Adjusts weights, not rules.

 Phase E — Meta-Learning & Integration

Learns how to learn. Adjusts learning parameters (not logic). Coordinates across all four preceding phases. Reward functions are Founder-defined only — the system cannot author its own goals. Topology-aware learning adapts based on structural context.

Adaptor Architecture — External Connections

Adaptors are external certified processes. Not part of the substrate. HTTP boundary only. The substrate never changes — only the adaptor configuration surface changes.

Adaptor Architecture — Burn Orchestrator, STABLE and VOLATILE registries

Stable Profile

  • 30–80 ms latency
  • 0.3% error rate
  • Predictable — deterministic behaviour

Volatile Profile

  • 20–500 ms latency
  • 5% error rate
  • Unpredictable — real-world variance

Same substrate. Same collapse pipeline. Same governance. Different constraint surface per vertical.

Vertical Regulation Status
FintechMiFID II, SOXProven
HealthcareHIPAA, FDA 21 CFR Part 11Proven
DefenceNATO STANAG 4586, MIL-STD-882EProven
InsuranceSolvency II, IDDProven
LegalEU AI Act (Art. 14), ECHR Art. 6Adaptor-ready
EnergyNERC CIP, IEC 62443Proven
AutomotiveISO 26262, SOTIF (ISO 21448)Proven
CybersecurityNIST CSF 2.0, SOC 2 Type IIProven
Supply ChainEU CSRD, CSDDD, Basel IIIProven
EdTechFERPA, COPPA, EU GDPR (Art. 22)Adaptor-ready
Governed AINo standard (substrate layer)Proven

Click any row to see how “Proven” is defined for that vertical.

Observability — Dual-Lane Cockpit

Two independent cockpits. Same KPI set. Different data sources. Zero shared state.

Stable Cockpit

Rate Surface
  • Current RPS
  • Min / Max / Avg
  • Historical Trend
Timing Surface
  • P50 / P95 / P99
  • Jitter
Reliability Surface
  • Error Rate / Count
  • Success Rate
Governance Surface
  • Throttle Rate
  • Capacity Utilisation
  • Accepted / Throttled

Muted · Precise · Minimal

Volatile Cockpit

Rate Surface
  • Current RPS
  • Min / Max / Avg
  • Historical Trend
Timing Surface
  • P50 / P95 / P99
  • Jitter
Reliability Surface
  • Error Rate / Count
  • Success Rate
Governance Surface
  • Throttle Rate
  • Capacity Utilisation
  • Accepted / Throttled

Expressive · Dynamic · Wide

Structural Properties

Property Enforcement
DeterministicSame input → same output. Always. No unseeded randomness.
Identity-freeNo personal identifiers in the substrate. Constitutional.
IrreversibleCollapse cannot be undone. C¹⁻ does not exist.
IdempotentC² = C. Collapsing a collapsed state changes nothing.
GovernedAll adaptation bounded by frozen constraint surface.
AdditiveEach neurotrophic phase adds capability. Nothing modified.
ConstitutionalDoctrine loaded at boot, immutable at runtime.

Execution Summary

Signal to resolution. One path.

1 Signal arrives
2 Connector validates (ingress boundary)
3 Atom dispatches (O(1) Instamap lookup)
4 Collapse pipeline executes (α → ν → λ → ρ → ε)
5 Admissible? — No → VETOED (structurally impossible)
6 Deterministic reaction produces output
7 Extinction returns substrate to purity (ω₀)
8 Telemetry emitted to neurotrophic layer (read-only)
9 Neurotrophic layer accumulates, learns, adapts (slow path)
10 Prediction confidence threshold met → adaptation surface mutates
11 Fast path continues. It doesn’t know anything changed.

The substrate never changes. Only the constraint surface changes.