| Market data · TradingView | Charts, indicators, real-time and historical data as a bot input. | Replace | TradingView licenses market data for display; its terms list automated trading and automated order generation among prohibited uses and exclude non-display use. [19] | Machine-readable data from a vendor or broker licensed for automated use. Keep TradingView as a human research surface. |
| Market data · MetaTrader 5 | Forex and CFDs, live or demo. | Keep | The official Python package connects to a running MT5 terminal and sends structured trade requests. [20] Prices are your broker's quotes, and demo fills are not live fills. | Record broker quotes, spreads and swaps as cost-model truth. Calibrate slippage on live fills only. |
| News & events | Economic calendar, news, sentiment. | Refine | Untrusted text that steers an LLM. Headline edits invisible to humans changed trading decisions in published tests. [13] | Quarantine, Unicode normalisation, hidden-text stripping, source provenance, arrival timestamps. |
| Alternative data | On-chain, macro, custom datasets. | Refine | Revisions, survivorship and licensing silently inflate backtests. | Store vintages as first published. Licence terms recorded per dataset. |
| AI orchestrator | Plan, reason, use tools, improve; three model vendors drawn in the box. | Refine | Orchestrating research is fine. Sending “approved strategies” onward to execution puts the learner in the trading seat. Hard-wiring vendors is a resilience risk. | Model-agnostic routing. Orchestrator coordinates research only; no broker credentials or order tools in reach. |
| Research agents | Market analyst, strategy researcher, code agent. | Keep | Hypothesis generation, literature synthesis and code drafting are where LLMs earn their place. [11] | Sandboxed, allowlisted egress, outputs are artifacts (notes, code, specs), never actions. |
| Research reviewer | Validate logic and find flaws. | Refine | A reviewer that shares the author's model, prompt and reasoning can share its blind spots, and words are not a test. | Independent reviewer (different model or prompt, no access to the author's rationale) plus automated leakage tests. |
| Approved strategies → execution | Orchestrator passes approved strategies to MT5 to place orders. | Replace | The drawing never says who approves. In a self-improving loop the answer cannot be the loop. | Named human promotes a signed version. Every order passes a deterministic risk engine before the broker adapter. |
| Order manager, broker adapter, execution monitor | Position, modify, cancel; live or paper; fills, slippage, errors. | Keep | Correct separation of concerns on the execution side. | Idempotent client order IDs, daily reconciliation against broker statements. |
| Memory & observability | Signals, orders, fills, PnL, market state. | Refine | Missing the one record that decides statistical honesty: every experiment ever run. | Experiment ledger with trial count; model, prompt, code and data hashes; hash-chained audit log. |
| Error analysis | Underperformance, drawdown, excessive trades, slippage, regime change, data issues. | Refine | Omits the failures that look like success: leakage, look-ahead, cost-model error, manipulated inputs, model nondeterminism. | Expanded taxonomy (module 10) with an owner per category. |
| Reasoning engine | analyze_failures() finds patterns in failed trades and proposes changes. | Refine | Pattern-hunting in a few losing trades is how noise becomes a rule. LLMs handle numerical time series poorly, by the account of the organiser of the 2025 live LLM trading contest. [9] | The LLM writes hypotheses in words, each with a mechanism. Statistics decide. |
| Validation & testing | Backtest, out-of-sample, transaction costs, risk limits, robustness. | Refine | Consulted every cycle, the out-of-sample set silently becomes in-sample. [4] | Purged cross-validation for development, a locked vault spent rarely, trial-adjusted Sharpe, then paper, shadow and canary. |
| Strategy update · Approve / Reject | Apply approved changes with version control (v2.1). | Replace | Approval is automated by the test result, so the system promotes itself. | Signed release, evidence pack, human approval, staged capital, rehearsed one-step rollback. |
| Loop step 5 · Compare | Keep if better, otherwise revert. | Replace | “Better” over a short window is a coin toss, and constant switching is itself overfitting that churns costs. | Champion and challenger on a pre-registered horizon, with a minimum track record and a significance test. |
| Deterministic risk engine & kill switch | Not drawn. | Add | The reference case: in 2012 Knight Capital sent more than 4 million orders in 45 minutes and eventually lost more than $460 million. [15] | Pre-trade limits, price collars, throttles, latched loss breaker, three-level kill switch (module 07). |
| Human promotion gate | Not drawn. | Add | Regulators expect the management body to stay accountable for decisions made by AI. [22] | Named approver, four-eyes deployment, evidence pack (module 09). |
| Security boundary | Not drawn. | Add | It combines private data, untrusted content and outbound channels (the lethal trifecta), then adds the power to trade. [17] | No path from untrusted text to an order (module 08). |
| Cost and capacity model | “Transaction costs” as one tick-box. | Add | An edge measured before spread, slippage and financing is not yet an edge. | Spread, commission, slippage, financing and market impact, calibrated on live fills (module 06). |
| Regulatory perimeter | Not drawn. | Add | Leverage caps, licensing and AI-marketing rules differ by jurisdiction and change. [24] [28] [29] | Perimeter review per jurisdiction, re-run on every material change (module 09). |
| Incident response & retirement | Not drawn. | Add | Published edges decay on average; the question is whether you notice yours first. [6] | Severity-graded runbook and retirement criteria written before launch (module 10). |