Recorded · practice only
Loading the saved case…
Recorded decision
Conditions set first
Full rules & risk limits
$10,000 paper account · 1% risk budget · 20% allocation cap · 1.5× ATR stop.
The original plan stays fixed as evidence arrives.
Evidence acquired
| Asset | 24h change | RSI 14 | ATR / price |
|---|
Select an asset for its original source. Missing values stay blank.
Agent’s full assessment
What changed
Original question
A claim is not permission.
Loading supporting evidence…
Ten unselected messages stayed pending. The controller required a named research check before the agent could act.
No selection, no research.
The code enforced research permission. This does not claim an LLM resisted ten jailbreaks.
Inspect all ten messages
Source failed. No trade.
Loading supporting evidence…
Four genuine RYO error responses from 20 October are re-evaluated through Ledger’s entry gate. No token measurements are invented.
No tradeResult of the offline check
0Positions opened
0Model calls
Retrospective check in temporary state, not a trade decision published during the outage. Original source dates are preserved.
Entry result & record chain
The no-trade event has a verified record chain. It has no assessment to replay.
Check the saved evidence.
Loading supporting evidence…
Recheck original bytes, recorded runs, fixed conditions and sizing using saved inputs and temporary state.
What verification establishes
- Playback shows the saved model events; it does not regenerate them.
- Recalculation checks the real case’s conditions and position size.
- Integrity checks saved run/capture hashes and the new outage event chain.
- Assessment replay applies only to assessed thesis events. The no-trade event is not labelled as a replayed assessment.
Live review records
No live thesis yet.
No thesis has been published in this book. Recorded cases are available in the walkthrough.
Review the market.
Checking connectionCollect current RYO evidence for BTC, ETH and SOL, then record a practice decision.
A fresh review uses the configured research services.
Estimated stagesEvidenceCompareDecideRecord
No run yet
Evaluation history & limitations
23 October: a relevant challenge opened ETH. The original pressure control failed and is preserved in the walkthrough’s recording history.
24 October: the scope guard held ten unselected messages pending with zero model calls. Selected research still opened ETH; incomplete cases waited.
The same-model chat baseline also opened a valid position. No reasoning advantage or independent-user comprehension benefit is established. The original trial used 57 model calls; the repaired trial used 22. Older attempts include tool-name errors. All are dated retrospective paper simulations with zero fees/slippage.