This page describes the compliance infrastructure voluntarily adopted by MorningEdge, currently under paper trading validation. MorningEdge is not a registered investment adviser or broker-dealer.
MorningEdge qualifies for the publisher's exclusion (Section 202(a)(11)(D) of the Investment Advisers Act of 1940), as affirmed by the Supreme Court in Lowe v. SEC, 472 U.S. 181 (1985). The Lowe Court held that publishers of impersonal commentary, distributed on a regular basis to the general public as a bona fide publication, are categorically excluded from the definition of "investment adviser." MorningEdge's content satisfies all three prongs: it is impersonal (no individualized recommendations), bona fide (published on a regular schedule as part of a genuine research program), and regular (daily scan results and diary entries on a fixed cadence). The compliance infrastructure described on this page was voluntarily self-assessed and adopted — referenced as a framework for responsible operation, not as a claim of certification — informed by SEC Rule 17a-4, SEC Rule 204-2, FINRA1 Regulatory Notice 15-09, ISO 27001 Clause 10.2, and the NIST AI Risk Management Framework.
◆Nature of Published Content
MorningEdge publishes research, educational content, and system validation data. The following table clarifies what MorningEdge does and does not publish.
What MorningEdge Publishes
- General market research on pre-market gap patterns
- Algorithmic scan results — impersonal, rule-based selections published to the general public
- A daily trading diary documenting system behavior during paper validation
- Lab reports: controlled experiments with in-sample/out-of-sample methodology
- Educational content about algorithmic trading system design
- Timestamped validation records via public Telegram channel (audit trail for system verification)
What MorningEdge Does Not Publish
- Personalized investment advice or individualized recommendations
- Recommendations to buy, sell, or hold any specific security for any specific person
- Portfolio management services or discretionary trading authority
- Solicitations for managed accounts or advisory relationships
- Guaranteed or implied returns — all published results are from simulated paper trading
- Content intended to direct any reader's specific investment decisions
◆Operating Principles
The system is organized around three principles that drive every architectural decision. For the full set of research-tested principles, see Principles.
Research-Driven
- Independent lab studies with in-sample / out-of-sample validation
- Deflated Sharpe Ratio correction for selection bias
- Knowledge base of published books and academic papers
Self-Auditable
- Hash-chained event ledger (tamper-evident, self-assessed)
- ISO 27001-informed corrective action register (not certified)
- Every parameter change traced to a lab finding
Paper Validation (Ongoing)
- Simulated paper trading since Day 0 — all results are from paper orders, not live capital
- Daily diary published with full trade details
- Shadow tracking on sit-out days proves discipline
◆Governance Framework
The table below maps each regulatory standard to the practices MorningEdge has voluntarily adopted — not as a claim of certification, SOC 2 equivalence, or independent audit, but as a self-assessed framework for responsible operation. None of these standards are binding on a publisher operating under Section 202(a)(11)(D). MorningEdge adopts them because an audit-ready posture is the minimum standard for operating an algorithmic research system. These are self-assessed practices, not third-party certifications.
| Framework | Implementation | Reference |
|---|---|---|
| SEC Publisher's Exclusion Lowe v. SEC, 1985 |
✓ Bona fide, impersonal, regular schedule | 472 U.S. 181 |
| NIST AI RMF 1.0 Model Risk Management |
✓ Model documentation, validation testing, human oversight (CHA model), ongoing monitoring — proportionate to system complexity | NIST AI RMF |
| ISO 27001 Clause 10.2 Corrective Actions |
✓ Finding → root cause → fix → evidence chain | ISO 27001 |
| Deflated Sharpe Ratio Lopez de Prado2 |
✓ All trials registered, selection bias corrected | SSRN 2460551 |
| Tamper-Evident Audit Trail Consolidated Audit Trail |
✓ Complete, timestamped, cryptographically chained record of all system decisions, deployments, and changes | Audit infrastructure |
| Data Integrity Survivorship Bias Control |
✓ [morningedge:data_tickers]+ tickers incl. delisted | Lab reports |
◆Conflict Mitigation
Conflicts of interest are addressed through three mechanisms: pre-publication validation records for all scan results, full disclosure of the operator's paper trading activity, and compartmentalization of AI provider access to proprietary strategy data.
Pre-Publication Validation
Every scan result is published to a public Telegram channel before market open as part of the 90-day paper trading validation. The public Telegram channel serves as a validation mechanism — it creates a timestamped, publicly accessible, timestamped archive allowing independent verification that selections were published pre-bell, not backfitted from results. The channel is operator-controlled, but Telegram timestamps are platform-generated and publicly verifiable by any reader. This allows any observer to independently verify that published diary results match the pre-market selections. All readers see the same scan results at the same time the system executes paper orders. No advance access, no preferential execution. During the current paper trading phase, all executions are simulated — no live capital is involved.
Conflict of Interest Disclosure
FINRA Regulatory Notice 15-09 identifies three questions relevant to conflicts of interest. MorningEdge addresses each directly:
- Does the operator trade the same securities? During the paper validation phase, the system executes simulated paper orders on the same selections published to the Telegram channel. The operator does not currently trade live capital on these selections. If and when live trading begins, the operator's orders will be submitted by the same automated script that publishes the validation record — eliminating any advance trading window.
- Is there preferential access or timing? No. All scan results are published simultaneously to a public channel. There is no tiered access, no early disclosure, and no preferential execution. Position sizes are determined algorithmically, not discretionally.
- Are there undisclosed financial incentives? No. MorningEdge does not currently receive compensation for publishing scan results. There are no affiliate arrangements, payment-for-order-flow agreements, or undisclosed financial relationships with any broker, data provider, or third party. Full disclosure is maintained on the disclaimer page.
AI Governance
Strategy parameters, position sizing, and exit rules are never sent to cloud AI providers. AI assists with research, literature review, and code — but proprietary trading logic stays local.
Zero strategy data in cloud AI
◆Human Accountability
Every decision that affects capital, risk parameters, or production deployment is made by a single human authority — the Chief Human Agent (CHA). Three AI systems build, advise, and analyze, but no AI agent has autonomous authority to deploy code, commit capital, or modify risk parameters. The CHA reviews and approves all production changes at Gate 5 of the research flywheel and all validation results at Gate 7. This is documented in the team structure and enforced by the corrective action register.
Decision Authority
Capital allocation, go/no-go on trades, risk parameter changes, and deployment approval — all require CHA sign-off. AI agents cannot bypass this gate.
AI Agent Boundaries
Each AI agent has a defined role, defined data access, and no ability to execute trades independently. The access control matrix is documented in the compliance manual, available on request.
Kill Switch
The CHA can halt all trading immediately via --close-all, which triggers a two-phase limit close on all open positions. Automated scheduling can be suspended in seconds via launchctl.
◆Fairness
All readers see the same scan results at the same time via a public Telegram channel that serves as a validation mechanism. There is no tiered access, no preferential execution, and no advance disclosure. The system's paper orders are executed by the same automated script that publishes the validation record — eliminating any window for selective trading. All scan results are published regardless of outcome; there is no selective reporting of winners.
Backtests include both active and delisted tickers to prevent survivorship bias — a form of unfairness to readers who would otherwise be shown inflated performance metrics based only on securities that survived.
◆Explainability
The production trading system is entirely rule-based — it is not a neural network, large language model, or black-box machine learning system. Every selection can be traced to specific filter criteria: pre-market gap percentage, relative volume, average daily volume, float, sector, and catalyst classification. These thresholds are documented, versioned, and testable. The ML gap predictor is a secondary research component and does not control production trade selection.
This design is intentional. Rule-based systems provide what Gagan3 Deep describes as "transparent, economically interpretable rules that can be validated against theoretical expectations and audited for regulatory compliance" (Walk-Forward Validation, p. 1). A regulator or auditor can inspect each filter threshold, understand why it exists (traced to a specific lab finding), and verify that it was applied consistently.
AI-washing defense: MorningEdge uses the term "algorithmic" precisely. The scanner applies deterministic, human-authored rules — not AI inference. AI tools (Claude, Gemini) assist with research, code review, and literature analysis, but they do not make trade selection decisions, generate signals, or sit in the execution path. References to "AI" in the team structure mean development assistants, not trading agents. This distinction matters: the SEC has cautioned against "AI-washing" — the practice of overstating AI capabilities to attract interest (SEC Staff Statement on AI, 2024). The system is rule-based, and is described as such.
◆Dataset
The system is validated against a dataset designed to eliminate survivorship bias and provide statistically meaningful sample sizes. It includes both active and delisted tickers to prevent the inflated performance metrics that result from testing only on securities that survived. All validation uses in-sample/out-of-sample splits to control for overfitting and selection bias — what Lopez de Prado identifies as "arguably the most fundamental question in quantitative finance" (Advances in Financial Machine Learning, p. 206) and what Chan describes as "almost impossible to completely eliminate… as long as one is building data-driven models" (Quantitative Trading, p. 75).
◆The Research Flywheel
Every change to the production system follows a gated research cycle with eight sequential gates — from diary observation through knowledge base research, controlled experimentation, gate review, and validated deployment. No gate can be skipped. No parameter reaches production without completing the full cycle. The gates are documented in the lab report registry and enforced by the corrective action register.
◆Evidence Chain
Every parameter in the production system traces to a specific lab finding through an unbroken chain of custody. The following example illustrates the full path from a live observation to a production deployment, including the intermediate artifacts that make it auditable.
Knowledge Base
Published books and academic papers ingested into a hybrid dense + BM25 retrieval system. Every research session draws on Kissell4, Carter5, Chan, Lopez de Prado, and more — ensuring decisions are grounded in published literature, not guesswork.
Hybrid search · literature-backed decisions
Corrective Action Register
Following ISO 27001 Clause 10.2: every finding is tracked through root cause analysis, corrective action, and verified evidence of effectiveness. Nothing gets swept under the rug.
ISO 27001 · full lifecycle tracking
Automated Scheduling
Scheduled jobs orchestrate the full trading day: draft scan, final scan with execution, regime recheck, shadow tracking, position close, safety net, and EOD tracking. Fully hands-off from open to close.
Automated Mon–Fri · open to close
Two-Phase Limit Close
Positions close via liquidity-aware limit orders first, falling back to market orders only as a safety net. Minimizes slippage on exit — the most overlooked cost in retail trading systems.
Limit-first · market safety net
Example: From Observation to Production
A diary entry flagged an anomaly in live results. The team searched the knowledge base for relevant literature, designed a controlled lab experiment with in-sample and out-of-sample splits, passed the gate review, and deployed the fix — with full traceability from observation to commit.
Telegram Validation Channel
Every scan result is published to a public Telegram channel before market open — creating a timestamped, publicly accessible, timestamped archive that proves selections were made pre-bell, not backfitted from historical data. The channel exists as a validation mechanism and audit trail: it allows any observer to independently verify that the system's published diary results match pre-market selections. This is how the system is validated in real time during the paper trading period. Each publication includes a disclaimer link. All readers see the same scan results at the same time — no preferential access.
Pre-market timestamps · independent verification · disclaimer on every publication
◆System Dependencies
The system depends on external providers for execution, data, and AI capabilities. Each dependency is documented with its failure impact and mitigation.
Broker: Alpaca
Paper trading execution via API. If the Alpaca API is unreachable at market open, no orders are submitted and the system logs the failure. Migration to Interactive Brokers (multi-broker architecture) is planned for production.
AI Providers
Claude (Anthropic) is the primary builder. Gemini (Google) advises on strategy. Neither is in the execution path — the trading system runs locally on deterministic rules. If all AI providers are unavailable, the scanner and executor continue to operate normally.
Data: FirstRate + Alpaca
Historical data (FirstRate) is stored locally. Real-time quotes (Alpaca) are required for live scanning. If real-time data is unavailable, the morning scan cannot run and no selections are generated.
Telegram
Validation records are published via the Telegram Bot API. If Telegram is unavailable, the system's paper orders are still executed but the public timestamp record is delayed. Local logs serve as the primary audit trail; the Telegram channel is a secondary, publicly verifiable record.
◆Data Retention
As a publisher operating under Section 202(a)(11)(D), MorningEdge is not subject to the recordkeeping requirements of SEC Rule 204-2 (investment advisers) or SEC Rule 17a-4 (broker-dealers). MorningEdge voluntarily retains trading records — orders, fills, scan history, regime decisions, exit signals, and shadow tracking — for a minimum of 10 years because thorough recordkeeping strengthens the credibility of published research. For reference, this exceeds the 5-year minimum under Rule 204-2 and the 6-year requirement under Rule 17a-4, though neither rule applies to MorningEdge's operations. Source code and compliance documents are retained indefinitely. AI conversation logs are retained for 5 years.
Critical CSV log files are hash-chained: each row includes a SHA-256 hash of the previous row, creating a tamper-evident record. Any inserted, deleted, or modified row breaks the chain. Planned infrastructure includes WORM (Write-Once-Read-Many) cloud storage via Wasabi Object Lock, where records cannot be deleted or modified by anyone — including the account owner — until the retention period expires. For reader and website data retention, see the privacy policy.
◆Validation Program
MorningEdge is currently in a 90-day paper trading validation phase. No live capital is at risk during this period. All published results reflect simulated paper orders executed through the Alpaca paper trading API — not real brokerage transactions. Every scan result is published to a public Telegram channel before market open, creating a timestamped public validation record. On days the regime filter signals YELLOW or RED, the system sits out — and runs shadow tracking to document what would have happened. This data is retained under the same hash-chain integrity controls as paper trade records.
◆Registration Status
MorningEdge is not registered with any federal or state regulatory body. The following table summarizes MorningEdge's registration status and the applicable exclusions relied upon.
Not Registered As
- Investment Adviser — not registered with the SEC or any state securities regulator
- Broker-Dealer — not registered with FINRA or the SEC
- Commodity Trading Advisor — not registered with the CFTC or NFA
- Transfer Agent, Municipal Advisor, or Funding Portal — none of these registrations apply
Applicable Exclusions
- Publisher's Exclusion — Section 202(a)(11)(D) of the Investment Advisers Act of 1940, as interpreted in Lowe v. SEC, 472 U.S. 181 (1985): publishers of impersonal, bona fide, regular commentary to the general public are excluded from the definition of "investment adviser"
- No advisory relationship — MorningEdge does not provide individualized advice, manage accounts, or accept discretionary authority
- No compensation for advice — no fees are charged for access to scan results or research content
◆Known Limitations
This page describes infrastructure that is under active development. The following limitations apply:
- All results are from simulated paper trading. No live capital has been deployed. Paper trading validates system logic but does not validate execution quality, market impact, slippage under real liquidity conditions, or the psychological and operational challenges of trading with real money. Past results, including paper trading results, are not indicative of future performance. Simulated trading programs have inherent limitations: no representation is being made that any account will or is likely to achieve profits or losses similar to those shown. While the system publishes selections before market open (eliminating hindsight in selection), paper fills do not reflect the market impact, partial fills, or queue priority that occur in live markets.
- All compliance claims on this page are self-assessed. No third-party audit or certification has been conducted. MorningEdge intends to pursue independent assessment before transitioning to live trading.
- The system is operated by a single principal (the CHA). This creates key-person risk that no amount of infrastructure fully mitigates. Business continuity planning for the production phase includes dedicated hardware, automated failsafes, and documented manual override procedures.
- The regime filter reduces but does not eliminate exposure to adverse market conditions. The catastrophic floor limits individual position losses but cannot prevent correlated drawdowns across all positions.
- WORM cloud storage, Ed25519 digital signatures, and the Raspberry Pi key vault are planned but not yet deployed. Current tamper-evidence relies on hash-chained CSV files stored locally.
- This system does not provide personalized investment advice, manage client assets, or accept discretionary authority over any account.
- State securities law. Certain states — including Texas (Texas Securities Act, Art. 581-4) and Massachusetts (MGL Chapter 110A) — may impose registration requirements on publishers that go beyond the federal publisher's exclusion. MorningEdge has not obtained state-level registrations or exemptions. MorningEdge is consulting with securities counsel on state registration requirements before any transition to live capital. Readers in jurisdictions with expansive definitions of "investment adviser" should consult their own legal counsel before relying on any published content.
- Reader data privacy. MorningEdge does not collect, store, or process personal financial information about readers. The Telegram validation channel is public and does not require readers to disclose personal information. Website analytics are limited to standard server logs. For full details, see the privacy policy.