Market Snapshot
Compressed 1h and 4h features that feed the scout agents.
| Symbol | Trend | ATR % | Volume ratio | Support | Resistance |
|---|---|---|---|---|---|
| BNBUSDT | +0.07 | 0.76 | 0.38 | 640.02 | 646.35 |
| BTCUSDT | +0.05 | 0.90 | 0.52 | 70163.0 | 71077.7 |
| DOGEUSDT | +0.22 | 1.49 | 0.47 | 0.09346 | 0.09494 |
| ETHUSDT | +0.26 | 1.28 | 0.81 | 2140.15 | 2168.0 |
| SOLUSDT | +0.51 | 1.19 | 1.15 | 89.3 | 90.78 |
Risk Engine
Before/after comparison to show how the deterministic layer adjusts model output.
- Before risk
BNBUSDT Short 10x, BTCUSDT Short 10x, DOGEUSDT Short 20x, ETHUSDT Skip, SOLUSDT Long 12x
- After risk
BNBUSDT Short 10x, BTCUSDT Short 10x, SOLUSDT Long 12x
- Reason
DOGEUSDT: Volatility budget is too high for this bootstrap run. ETHUSDT: M2.7 did not approve a position.
Scout Outputs
Two cheap models provide short JSON summaries. This page shows the human-readable slice.
| Symbol | Agent A | Agent B | M2.7 |
|---|---|---|---|
| BNBUSDT | Short 10x | ||
| BTCUSDT | Short 10x | ||
| DOGEUSDT | Short 20x | ||
| ETHUSDT | Skip | ||
| SOLUSDT | Long 12x |
Execution Timeline
A compact view of what happened from trigger to paper execution.
- 20:52:38Building Features
Compressed 1h and 4h candles into feature bundles.
- 20:54:08Debating
Qwen scout produced a single low-token scan across the tracked symbols.
- 20:54:37Judging
MiniMax M2.7 judged the Qwen scout summary and proposed the final portfolio.
- 20:54:38Risk Checking
Risk engine applied disagreement, volatility, and capacity gates.
- 20:54:38Executing
Observe-only mode kept the run in planning state.
- 20:54:38Monitoring
Monitoring lane primed with current mark-price context.
- 20:54:38Reviewing
Review lane reserved for Claude-based post-run analysis.
- 20:54:38Reviewing
Created a Qwen scout + M2.7 judged plan from live Binance market data. ETHUSDT: M2.7 did not approve a position. DOGEUSDT: Volatility budget is too high for this bootstrap run.