How the pieces connect
Six layers, each with one job. The risk engine sits between every signal and every decision, and the record of what happened is kept separately from the thing that produced the signal.
From ingestion to review
-
0x01
Ingestion
Market data for each asset class is collected and aligned to common timeframes.
- Holds
- Time-aligned price, volume, flow and yield series
- Feeds
- Indicator layer
-
0x02
Indicator layer
Standard indicators are computed per instrument and per timeframe. Nothing is interpreted yet.
- Holds
- MACD, RSI, ATR, ADX and related readings
- Feeds
- AI research layer
-
0x03
AI research layer
One layer reads all three asset classes together, produces a written interpretation and grades any signal.
- Holds
- Interpretations, regime labels, graded signals
- Feeds
- Risk engine
-
0x04
Risk engine
Rule-based checks on volatility, liquidity and exposure. It can pass, reduce or reject a signal, and it sets position size.
- Holds
- Risk limits and current portfolio exposure
- Feeds
- Human decision
-
0x05
Decision and execution
A person decides whether to act. If a trade is taken, its entry, exit and indicator state are recorded.
- Holds
- Trade records
- Feeds
- Review
-
0x06
Review and journal
Closed trades are analysed and published. Each record carries a SHA-256 fingerprint so later changes are detectable.
- Holds
- Journal entries and their fingerprints
- Feeds
- Back into research
One research layer, one risk engine
Signals from different markets are not sized in isolation. They meet in a single risk engine that sees total exposure.
Why records are fingerprinted
A trade journal is only worth reading if its entries cannot be quietly edited after the fact. Each published record is reduced to a fixed JSON form and hashed with SHA-256. The hash is shown beside the record.
Anyone can recompute the hash from the published figures. If a figure changes, the hash changes with it.