# Market Intelligence & Autonomous Validation Engine
## Single Source of Truth Product + EPIC Specification — v3.0

**Date:** 25 August 2026  
**Status:** Authoritative single source of truth — complete internal-production specification  
**Purpose:** Closed internal-use market intelligence, recommendation delivery, realistic paper-portfolio simulation, continuous validation and gated agentic improvement. Not commercialized at this stage.

> The full authoritative v3.0 specification is governed by this repository. Product implementation MUST satisfy the requirement registry and traceability gate. The baseline product definition, architecture, strategies, immutable ledger, agent loop, simulation, pre-open/post-close executive emails, UI, observability, security, data quality, replay, and promotion rules are mandatory.

## Governing principles
1. Market isolation: TASI and S&P 500 are never mixed in one ranking.
2. Multiple strategies: no universal stock score; strategies can disagree.
3. Point-in-time truth: only information available at the decision timestamp may be used.
4. Immutable predictions: recommendations/evidence are never rewritten after outcomes.
5. Score is separate from confidence.
6. Explainability: expose factors, evidence, contradictions, target method and invalidation.
7. Simulation before capital: prove via virtual portfolios and benchmarks.
8. No uncontrolled self-modification: agents propose; validator/promotion gate decides.
9. Reuse before rebuild: mature OSS first; `tasi.py` is optional.
10. Deterministic facts: LLMs classify/reason; numeric facts come from timestamped sources/calculations.

## Initial markets
- TASI Main Market — primary target.
- S&P 500 — developed-market control.

## Required product loop
Data Agent → Analyzer → Risk/Portfolio Agent → Simulator → market outcome → Observer → Diagnoser → next-day Analyzer context → Research/Challenger → Validator → Promotion Gate.

## Required daily delivery
- Pre-open luxury executive HTML recommendation email exactly two hours before the enabled market opens.
- Post-close luxury executive HTML accountability email after verified close data, evaluating the immutable morning recommendations as RIGHT / WRONG / STILL OPEN with reasons and rule/evidence traceability.

## Required simulation behavior
The user supplies a virtual budget and instructs the system to treat it as real money. The simulator may BUY, HOLD, ADD, TRIM, SELL, SKIP, or deliberately DO NOTHING under explicit RuleRegistry constraints. No real brokerage execution is allowed in this stage.

## Required auditability
Every recommendation, action, and deliberate non-action records the reason, cited rule IDs and versions, engine version, evidence references, data snapshot, portfolio state, timestamps, and subsequent outcomes. Corrections are append-only.

## Required UX
Graphical, executive, emotionally legible, simple for users without a financial background, with expert evidence one level deeper. Market selection changes the complete analytical context; markets are not mixed in ranking tables.

## Required implementation discipline
All mandatory requirements are represented in `requirements/registry.yaml`. A requirement is not complete until it has implementation and acceptance-test traceability. CI blocks releases with missing mandatory traceability.

## Scope note
This repository's full SSOT is versioned. Any change to scope must be made as an explicit SSOT revision; implementation notes, issues, prompts, or agent decisions cannot silently change product scope.