Trade & decision journal
The immutable, frozen-at-T record of every observation, proposal and trade — the heart of the research.
Without it, weeks of research teach nothing. Every observation, proposal and trade is frozen at time T and immutable. Only outcome and lessons learned are filled in later, and they point to a frozen record without modifying it. Nothing is forgotten.
The heart of the research
The journal is what lets you ask the only question that matters a month from now: was there a real edge, or just narration? It feeds the memory layer and the continuous-learning loop.
Schema
| Field | Type / values | Note |
|---|---|---|
id | id | |
profile | string / enum | Profile / asset-class discriminator (e.g. crypto, forex) |
timestamp | datetime (UTC) | Exact instant, never modified |
type | SHADOW_OBSERVATION · PROPOSED · APPROVED · EXECUTED · REJECTED · MISSED · EXIT | Includes shadow observations and opportunities seen but not taken |
symbol | string | Token concerned |
market_snapshot | json (frozen) | Price, indicators that mattered, volume / liquidity at time T |
evidence | json (frozen) | Quant, on-chain and social evidence the agents relied on |
scores | json | Signal · AI · risk · memory · strategy · execution · final |
decision | buy · hold · reduce · exit · skip | |
reasoning | string (frozen at T) | Never regenerated later |
governance | AUTO · HUMAN | Acted within permissions vs routed to validation |
human_verdict | json (if HUMAN) | Governor decision + their reason |
simulated_fill | json | Execution price after slippage + fees |
position_ref | id | Link to the position |
outcome | json (filled later) | PnL, duration, exit trigger |
lessons | json (filled later) | Lessons learned and future improvements |
Reasoning is frozen, never regenerated
The reasoning string is captured at the instant of the decision and is never rewritten after the fact. That immutability is the entire point: hindsight cannot leak into the record.
What it unlocks
A month from now you can filter “all shadow observations and entries, their reasoning and evidence at the time, and the outcome” — and discover which patterns actually paid off before any capital was at risk. The journal is the raw material for knowledge discovery and the learning loop.
Open question: how much to log
Does the lab continuously log opportunities seen but rejected (MISSED) from the start, or only active decisions, to avoid drowning the journal? Tracked on the roadmap.