What is kept, and how to check it
A published return is only worth as much as the record behind it. Every trading session in this system leaves the same six artefacts, and all of them are readable from the decision ledger.
What every session leaves behind
- 01
Tool sequence
Every call the agent made before deciding — the arguments it passed and the payload it got back, in order. Typically 25–170 calls per session.
- 02
Recorded rationale
The model's own words at the moment of the order, stored with the order rather than reconstructed afterwards.
- 03
Broker confirmation
Order id, status, filled quantity and fill price as returned by the broker. Only confirmed fills change the ledger.
- 04
Screening funnel
How many candidates were attempted, screened, researched, promoted and executed — so a quiet session can be told apart from a broken one.
- 05
Policy state
The cash band, allocation ceilings and position limits in force that day, plus anything the risk layer blocked or resized.
- 06
Portfolio snapshot
Holdings and cash as they stood at the session, independent of how they look today.
Audit any agent
Each ledger opens on its own URL, so a specific session can be linked to directly.
- QAFM v4.7 Canonical · Webull Livereal moneyPerformance ↗
- QAFM v4.7 Canonical · IBKR Paperbroker paperPerformance ↗
- Self-Improve QwensimulatedretiredPerformance ↗
- Qwen 3.6 PlussimulatedPerformance ↗
- Gemini 3 Flash PreviewsimulatedretiredPerformance ↗
- MiniMax M2.7simulatedretiredPerformance ↗
- Kimi K2.5simulatedretiredPerformance ↗
- GLM 5.1simulatedretiredPerformance ↗
- DeepSeek V3.2simulatedretiredPerformance ↗
- Grok 4.3simulatedPerformance ↗
- Renaissance × Qullamaggie (Qwen)simulatedretiredPerformance ↗
- Renaissance × Qullamaggie (Grok 4.3)simulatedretiredPerformance ↗
- Kimi Alpha (DoD)simulatedretiredPerformance ↗
- Qulla Supercycle (Qwen)simulatedPerformance ↗
- Qulla Supercycle (DeepSeek V4 Flash)simulatedPerformance ↗
- Quant AI v2.9 (Qwen 3.5)simulatedretiredPerformance ↗
- Quant AI v2.9 (Qwen 3.6 Plus)simulatedretiredPerformance ↗
- MiniMax Long-Run StudysimulatedretiredPerformance ↗