Echo Module (EM)

Charter · Version 1.1 · Draft · Owner: Jean-Baptiste Decorzent
STATUS: DRAFT — PENDING APPROVAL
Table of contents

1 · Purpose & Definition

This Charter defines what Echo Module is, what it is not, who owns it, and what rules govern its construction. It is the reference document every later decision — strategic, architectural, or implementation-level — is checked against. Nothing proceeds to Phase 2 (Strategy) until this Charter is approved.

Definition
Echo Module is a domain-agnostic engine that fuses unreliable, multi-source signal into an explainable, calibrated, provenance-tracked state — informing a decision, never autonomously acting on it.

EM is not

EM is infrastructure. It has no opinion about mental health, careers, or markets. It only knows how to take signals of declared reliability, fuse them honestly, learn which ones to trust over time, explain why it concluded what it concluded, and stop short of acting on its own conclusion.

2 · Origin & Evidence Base

EM is not a speculative abstraction. It is the documented, convergent result of the same architecture being independently built five times, in five unrelated domains, by the same author, without the abstraction being named until now.

SystemDomainEvidence artifact
haiva — GuardianClinical mental healthsignal_fusion.py — weighted fusion of NLP + VoiceBiota + physical biomarkers into an auditable, deterministic patient state (NONE→CRITICAL)
sona — Living NarrativePatient AI companionWeekly synthesis of accumulated conversation into a coherent, evolving narrative of the person
nara — Signal RouterConsumer wellbeingTiered escalation (Haiku/Sonnet) by risk level — routing cost/rigor to stakes, mirrored in TRAXON's conviction-gated LLM debate
LWE — BackboneCareer intelligenceProvenance-tracked, confidence-scored knowledge graph; raw layer never overwritten; confidence evolves 0.4→1.0 with corroborating evidence
TRAXON — Conviction EngineCrypto-derivatives tradingfinal_score = Σ(score × eff_weight/available_weight) — renormalized weighted fusion across 32 sources, plus a formal calibration loop: predictive_weight = clamp(1.0 + 0.3×(accuracy₉₀ᵈ − 0.5), 0.30, 1.50)
Why this matters for the Charter
TRAXON's formulas are the most mathematically explicit version of what the other four systems do informally or qualitatively. This Charter treats TRAXON's conviction-engine math as the closest thing EM has to a reference specification — not because trading is more important, but because it is the most rigorously falsifiable implementation of the shared pattern.

3 · Scope Definition

In scope — the six core components

ComponentDefinition
SourceA named signal producer: fetches a raw value, scores it to a normalized [0,1] range, declares a base weight and a fallback value for when it is unavailable.
FuserCombines available Source outputs into one state via weighted average renormalized over available sources; includes incongruence/independence detection to flag when sources disagree or when a source is statistically uninformative.
CalibratorAdjusts each Source's effective weight over time based on verified accuracy against real outcomes; blends prior belief with live evidence weighted by sample size.
ProvenanceStoreRecords which Sources contributed to which fused state, their value at the time, and — once known — whether they were correct. The substrate the Calibrator and any audit trail depend on.
GateConverts a fused state into a discrete classification against thresholds, and enforces the inform-vs-act boundary: EM outputs a state and a confidence level; a separate, explicit decision — outside EM — decides what happens next.
RouterEscalates computational cost/rigor in proportion to stakes or uncertainty, or conditionally triggers a second, more expensive opinion gated on cost/ambiguity. Scope is deliberately narrow — see resolution note below.
Resolved by Phase 1 audit — promoted with a scoped definition
Router accumulated six structurally distinct candidate shapes during the audit. Only two earned independent, cross-domain confirmation strong enough for core status: stakes-proportional escalation (3 independent systems) and cost-gated second opinion (2 independent + 1 corroborating fork). The other four shapes — reliability-fallback-on-failure, static hardcoded tiers, human-selected/persisted-preference, and cadence-based tiering — are real but single-system evidence; they ship as documented candidate variants, not as confirmed core behavior, until a second independent instance of each is found. Full evidence in EM_Candidate_Matrix.md, §2.6.

Explicitly out of scope

BOUNDARY TEST — applied to every design decision:
"If this cannot be justified without naming haiva, sona, nara, LWE, or TRAXON specifically, it does not belong in EM."

4 · Access Model

Stated as directional intent. Sequencing, licensing terms, and monetization are Strategy Plan decisions (Phase 2), not Charter decisions.

LayerFormPurpose
1Open-source core libraryZero-PII, domain-agnostic fusion/calibration/provenance logic. Builds credibility and reusable evidence of correctness. Natural fit for infrastructure-focused open-source grants.
2Hosted APIManaged service built on the open core, for teams who don't want to run it themselves.
3Embedded / commercial licenseSelf-hosted licensed deployment for organizations that cannot send data to a third party.

5 · Governing Principles

Binding. Every milestone in the Development & Implementation Plan (Phase 4) is checked against these before being marked complete.

6 · IP, Ownership & Entity Status

Current state, as of this Charter
Jean-Baptiste Decorzent is the 100% personal owner of haiva, sona, nara, LWE, and TRAXON. None of the five currently sits under a dedicated incorporated entity built for it. EM, as of this Charter, is likewise personally owned by JB, with no entity assigned.

POLEarth SA is a separate, pre-existing entity in which JB holds director and signature rights, not shareholding. It was previously identified as a possible temporary vehicle relevant to LWE. Discussions with POLEarth's shareholders about EM — or about any of the five source projects — have not yet begun. This is a deliberate, considered choice, not an oversight: JB has assessed that having something demonstrably solid to show before opening that conversation is the stronger position, and has flagged confidence that POLEarth remains usable later if it proves the right vehicle. This Charter records that reasoning rather than resolving it — the entity decision is explicitly deferred, see below.

Provenance policy — clean-room extraction

EM is built by describing the pattern and interface found in haiva, sona, nara, LWE, and TRAXON — not by copying code out of those repos into EM. Where the Phase 1 audit identifies a reusable pattern, the resulting EM implementation is written fresh against the abstracted specification. Every EM module carries a provenance note identifying it as independently authored, informed by patterns observed in JB's own prior, solely-authored work.

Rationale: this keeps haiva/sona/nara/LWE/TRAXON's own IP position clean and unaffected by whatever EM becomes — including if any of them is later sold, licensed, or restructured independently of EM.

Entity decision — explicitly deferred

No decision is made in this Charter about which legal entity, if any, should ultimately hold EM (a new dedicated entity, POLEarth SA, personal ownership, or a non-profit/foundation structure — relevant if EM pursues open-source infrastructure grants where a commercial-entity applicant may not be the right fit). This question is revisited in the Strategy Plan, after the Phase 1 audit produces a concrete picture of what EM actually contains and what it's worth.

7 · Relationship to Source Projects

8 · Success Criteria — This Phase

  1. This Charter is reviewed and approved by JB.
  2. The companion Claude Code Audit Directive is executed, read-only, across all five source projects.
  3. A consolidated EM Candidate Matrix is produced and reviewed before Strategy Plan drafting begins.

9 · Governance & Amendment

JB is the sole approving authority for this Charter and any amendment to it, consistent with the ownership position recorded in Section 6. Material changes to EM's scope (Section 3) require an explicit, versioned Charter revision — not an informal decision made mid-build.

VersionDateStatusNotes
1.0September 2026Draft — pending approvalInitial Charter, drafted following identification of the shared signal-fusion pattern across haiva, sona, nara, LWE, TRAXON.
1.1September 2026Draft — pending approvalRouter promoted to sixth core component, scoped to the two independently-confirmed shapes, following completion of the Phase 1 audit (five Claude Code sessions across all source repos). Full evidence in EM_Candidate_Matrix.md.

Echo Module Charter v1.1 · Draft · Not for external distribution until approved