# NoahAI Technical Whitepaper

Version: v2.1 web source (PDF v1.9 retained as a snapshot)
Updated: 2026-08-04
Type: Public technical baseline

## 0. Executive Summary

NoahAI is AI financial decision infrastructure that connects users' financial knowledge and market data to judgment, validation, execution, and auditable records. Its flagship AI Custom product is an AI strategy operating system, not an auto-trading promise engine.

Its operating principles are:

- explainable judgment,
- clear responsibility boundaries,
- replayable logs,
- risk-first execution control.

## 1. Scope and Purpose

This document defines NoahAI's public technical baseline.

Included:

- architecture and operating layers,
- judgment/execution separation,
- logging, XAI, and replay model,
- security and governance,
- KPI framework,
- recent operational updates.

Excluded:

- return guarantees,
- performance marketing claims,
- internal sensitive controls.

## 2. Problem Statement

Financial decision environments repeatedly face:

- information asymmetry,
- continuous monitoring burden,
- emotional bias,
- lack of auditable decision records.

NoahAI addresses this by structuring decision environments rather than giving opaque one-shot answers.

## 3. Identity and Responsibility Boundary

NoahAI is not:

- an asset custodian,
- a discretionary manager,
- an investment advisor.

NoahAI provides judgment assistance and risk controls.
Execution and fund movement are performed by user-controlled accounts and external exchange/broker APIs.

## 4. Architecture

NoahAI uses a 7-layer architecture:

1. Market Data Layer
2. Account State Layer
3. Decision Layer
4. Risk & Guardrails Layer
5. Execution Bridge Layer
6. Logging & Report Layer
7. Feedback Loop Layer

This layered model separates concerns and supports operational traceability.

## 5. Judgment/Execution Separation

- AI generates decision candidates and rationale.
- Execution occurs only under user policy and external API constraints.
- Success/failure/blocked outcomes are all logged.

This enables:

- explainability,
- auditability,
- reproducibility.

## 6. Data and Logging Model

Core event objects include:

- DecisionLog,
- MarketSnapshot,
- RiskEvent,
- ExecutionResult,
- XAITrace.

Data principles:

- minimum necessary collection,
- separation between personal and operational data,
- retention and anonymization policy,
- separation of source logs and aggregate reporting.

## 7. XAI and Replayability

XAI is treated as an operational accountability requirement.

NoahAI provides:

- human-readable rationale for decisions,
- post-hoc traceability,
- version-tagged policy comparison.

Replayability is supported through input-state logging and versioned decision context.

## 8. Security and Compliance

Security baseline:

- user assets/API keys remain user-controlled,
- minimized server-side data footprint,
- access separation and audit logs,
- masking and anonymization where required.

Compliance baseline:

- non-discretionary/non-custodial boundary disclosure,
- aligned disclosure and consent flows,
- explicit consent evidence for anonymized operational analytics.

## 9. Operational KPI Framework

KPI is used as operational health instrumentation.

Examples by layer:

- Auth: login/session/health-check events,
- Decision: inference completion and latency quality,
- Execution: order success/failure taxonomy and holding metrics,
- Operations: deployment reflection time and incident recovery.

KPI changes require synchronized update across collection schema, server aggregation, and UI.

## 10. Recent Baseline Updates (2026-07)

- Auth portal policy page and onboarding path alignment,
- separated consent items at signup with evidence storage,
- one-command/manual deployment standardization,
- update UX and exchange verification guidance improvements.

### v3.9.0.3 AI Custom and multi-AI architecture

- configurable indicator periods, timeframes, AND/OR rules, and requested-versus-calculated values in AI Custom advanced mode,
- per-task `{provider, model}` routing for OpenAI, DeepSeek, Claude, and Gemini,
- Kimi K3 and K2.6 as assistant trials until real-key validation,
- independent OpenAI transcription profile,
- model lifecycle, capability, and account-model validation,
- explicit separation of visible/learning sources from live-order targets,
- separation of AlphaArena comparison/replay from Client live-operation scope,
- full source regression: 1,135 passed, 6 skipped, 0 failed.

Windows v3.9.0.2 remains the user distribution. The v3.9.0.3 candidate was not separately distributed and is superseded by the v3.9.0.4 candidate below.

### v3.9.0.4 AI Custom roles and shared execution contract

- General mode confirms NoahAI-selected candidates with AI Custom exactly once. Advanced mode lets `StrategyUniversePolicy` and the original user strategy independently determine candidates, direction, TP/SL, and exits.
- Advanced mode does not require base-AI agreement. NoahAI does not silently rewrite strategy rules; a risk conflict reduces quantity or blocks the order.
- Each exchange and broker independently validates its supported universe, liquidity, spread, and data quality before producing the shared `TradeCandidate` and `ExitPolicy`.
- Regime scope is explicit: market, symbol, both, or none. Missing data does not silently switch scope.
- Users choose parallel, total-risk split, or priority-one execution across selected venues. Valid same-symbol executions share aggregate opportunity risk and results; only a repeated same-venue, account, and signal submission is blocked as a duplicate.
- LEARNING runs the complete decision and evidence pipeline while blocking external account-state changes and new orders. PAPER and LIVE use the same decision contract.
- Full regression is 1,219 passed, 6 skipped, and 0 failed. Windows build, signing, installation/update recovery, real connectivity, and long-run E2E remain distribution gates.

The detailed public contract is `/source/noahai-platform-v3904.md`.

### v3.9.0.6 AI strategy operating system and strategy-hub business layers

- NoahAI is defined as AI financial decision infrastructure connecting user knowledge and market data to judgment, validation, execution, and records.
- AI Custom connects `Learn → Create → Validate → Operate → Improve → Share` as an AI strategy operating system.
- Source Candidate 2 implements the AI mentor, source-to-rule tracing, advanced order plans, walk-forward and PAPER validation, and shared order/authenticated-fill contracts.
- The daltrading strategy hub uses nine purpose-specific categories and a default 30-day free beta. It does not fabricate listings when no strategy passes publication and evidence gates.
- Current client licenses, the free strategy hub, later creator subscriptions/marketplace, and B2B/IP/white-label offers are separate business layers. A paid strategy market is not promoted before payment, settlement, refunds, disputes, moderation, rights, and legal controls are ready.
- Full regression is 1,332 passed, 6 skipped, and 0 failed. The public Windows installer is v3.9.0.5 Fix Patch 1; v3.9.0.6 remains pending Windows rebuild, live-account checks, and long-run E2E.

NoahAI does not claim to replace TradingView's complete charting and Pine ecosystem. It focuses on the higher operating layer that validates, executes, and supervises knowledge from TradingView, books, videos, and users.

## 11. Enterprise Adoption View

Key enterprise evaluation points:

- clear responsibility boundaries,
- audit/replay readiness,
- security control and access separation,
- standardized deployment and recovery,
- explainable decision records.

## 12. Limits and Roadmap

Current limits:

- adapter complexity across heterogeneous APIs,
- environment-level performance variance,
- conservative rollout policy for advanced learning features.
- full Pine/arbitrary user-code execution and AlphaArena multi-engine live connectivity remain follow-up work.

Roadmap:

- stronger adapter verification automation,
- tighter KPI disclosure alignment,
- end-to-end traceability across policy, consent, and deployment logs.
- partial take-profit, trailing stop, re-entry, TradingView webhooks, and a user-indicator sandbox.

## 13. Conclusion

NoahAI's core value is not short-term return claims.
It is operationally reliable AI decision infrastructure with:

- safety,
- explainability,
- replayability,
- clear responsibility boundaries.

## 14. Public Documentation Policy

- Public technical documents do not expose internal/private paths.
- Sensitive operational details remain in internal-only documentation.
- Every public revision must update version/date/change baseline.