Your edge, under observation.
ZEC perpetuals on Aster. Every trade measured, every cost visible.
—
USDTLet the market
make its case.
Start a session to let your OpenRouter model evaluate live ZEC signals and explain its decisions here.
Market watch EVENT-DRIVEN AI
—Waiting for an adaptive watch plan.
Prices, P&L, stops and liquidation are continuously monitored without an AI call. Triggers can wake the AI before its deadline.
Shadow performance
ACTUAL ROLLING P&LMark-equity change minus reported AI spend. Short sessions show only their observed history, not a full hour or day.
Hourly / 3-hourly / daily run-rate preview
Open position 0
The bot will enter when its analysis clears your conviction threshold.
What’s happening
Your research trail begins with the first analysis.
The true cost
TRANSPARENT- Trading & liquidation fees
- —
- Funding paid / (received)
- —
- OpenRouter spend REAL USD
- —
- Mark P&L after AI spend
- —
Trading capital is virtual. OpenRouter calls use real credits. Open-position equity uses mark price; closing also incurs spread, impact and fees.
Every decision. Every outcome.
No hidden losses. No reset history.
Completed trades
NET OF COSTSDecision journal
Latest 40Session archive
Learning, with a paper trail.
Three signal families. Adaptive weights. No promises of alpha.
How the decision is made
ENSEMBLE V1Observe
Live Aster mark and depth, with completed 1m, 5m and 15m candles.
Evaluate
OpenRouter weighs trend, breakout and reversion evidence and explains its conviction.
Measure
Simulated book fills, fees, funding and maintenance-aware exits determine net results.
Adapt
Completed-trade margin returns adjust expert weights, shrunk toward a 20-trade prior.
What “learning” means here
This is online strategy weighting and outcome feedback to the AI, not autonomous code rewriting or proof of future profitability. Learning starts fresh in each session. The first 20 outcomes per expert are labeled as limited evidence; even larger samples can overfit. Strategies are not randomized or controlled experiments. Weights change the AI’s evidence—not your leverage ceiling.
Execution model & limitations
USDT linear perpetuals, one isolated ZEC position. Entries walk up to 20 observed book levels plus your extra slippage. Exits may extrapolate beyond depth, explicitly flagged. Fees are charged on each fill’s notional. Liquidation uses mark price, venue maintenance tiers when available, and a conservative closing-fee reserve; full liquidation is simulated, not Aster’s partial-liquidation or ADL engine. Losses are capped at isolated collateral plus entry fees; a gap deficit is shown as an insurance credit.
Published funding is reconciled once, including trades closed before publication. If settlement mark is unavailable, the observed mark is an estimate. Market outages may hide intervening stops or liquidations and are flagged. Latency, changing depth and exchange-specific behavior mean shadow results will differ from actual fills. There is no guaranteed profitable strategy.
Set the conditions.
Changes apply to future entries. Open positions retain their execution settings.
Connection status
Your existing OpenRouter key stays on the server. Aster signing credentials, if configured, are used only to read fee and maintenance rules. This app cannot submit exchange orders.
Jev P&L & performance.
The Laya-based, Jev-inspired engine—not official Jev. Attributed trading results and operational evidence.
Jev net P&L
Jev-attributed completed trades
Connecting…
Waiting for authenticated engine status.
Regime signal. Mechanical risk.
Uptrend → long; downtrend → short; mixed → hold. Requires ≥55% semantic support, a ≥10-point lead and agreement with trend signals. Not a calibrated win probability. Cost and risk checks can reject any entry.
Loading evidence…
Response latency
END TO ENDLatest 60 decisions, oldest to newest. Response times include cold calls; failed calls have no recorded duration.
Decision outcomes
NOT TRADE RESULTSA valid response may still be rejected or expire. “Observed” means recorded without execution authority.
Forward evidence
MARK LABELS ONLYThese labels track the subsequent ZEC mark-price change, not strategy returns. Gap-flagged and missing observations cannot establish an edge. Executable prices, funding-adjusted returns and probability calibration are unavailable.
Model cohorts
SEPARATE VERSIONSVersions and policies are separated below. Providers observe selective primary-analysis events; these are not matched, randomized comparisons.
Decision history
LATEST FIRSTExpand a decision for semantic scores and 15-minute / 1-hour observations. A semantic likelihood is not the probability of a profitable trade.
Compare one change.
Independent virtual books on one prospective local signal stream. Proposed defaults are experimental, not evidence of an edge.
Readiness and current cohort
Open this page to load authenticated experiment status.
Pause primary entries and turn the optional observer off yourself before enabling. This substitutes the new common-sampling experiment for primary entry sampling; existing protective exits continue. No paid fallback or real orders.
Capacity diagnostics unavailable. Forecasts are not a capacity guarantee.
Register a reviewed manifest and storage plan
Prepare both JSON files locally from the downloaded status export. Paste the exact manifest and a separate forecast-and-stop storage plan with explicit source, execution and headroom estimates. Forecasts are not a capacity guarantee. Registration stores both immutably and starts disabled; it does not raise the configured global trace cap, activate a cohort or reset an account.
No manifest reviewed.
No storage plan reviewed. Forecasts are not a capacity guarantee.
Current results and latest hourly record
Costs are already included in net. Unknown coverage and infrastructure cost stay visible. Hourly monitoring never retunes, declares a winner or promotes real trading.
Current cumulative report
No cohort registered.
Latest durable hourly report
No hourly record yet.