Date: 2026-11-03. State: Accepted for paper-first by the answer to Q3; technical details proposed.
Context#
Automated execution needs evidence of correctness and recovery. The user confirmed backtesting and paper trading in the MVP in Q3.
Decision#
MVP backtest + paper, without a live adapter enabled. Use deterministic strategies and risk; an LLM, if introduced in the future, only produces reviewable configuration.
Alternatives#
Live from the start, or an LLM with direct access to orders. The first expands the scope criteria; the second does not preserve the proposed deterministic engine.
Consequences#
Allows testing contracts without real capital, but paper does not prove profit nor liquidity fidelity. A paper broker is still a later decision after the market.
Review trigger#
Future explicit scope change, with limits, evidence and its own operational authorization.