Broker equity snapshots during dashboard refreshes. Focuses on real marks, not overnight fantasy.
The edge it is pursuing, the limits it cannot waive, and how it learns without grading its own homework.
The goal is to beat the S&P 500. Outperformance is not guaranteed, and early results are not proof of durable edge.
Every strong candidate is scored later against SPY whether traded or skipped, so the desk learns selection quality instead of grading its own homework.
7 candidate signals logged · 1 traded · 6 skipped.
What is still open and can still change the final result. Marks come from the paper broker when available.
Each holding uses its raw broker or market quote, never a made-up scaled price. Equity is broker-synced so P/L stays broker-true.
Supervised live-paper smoke test: single-share AAPL bracket to verify place + readback + journal path on dedicated paper account only.
Alibaba/ByteDance/DeepSeek may soon receive H200 chip approval (positive export catalyst); most-active volume and +2.2% price action confirm near-term technical strength.
grok-4.5 and openai/gpt-5.6-sol independently review the same evidence without seeing each other's answer. If the action or symbol differs, nothing executes. Every blocked disagreement and unavailable review appears here.
Orders require independent consensus. Disagreement, timeout, malformed output, or low confidence blocks execution with no fallback.
| When | Block reason | Primary | Reviewer |
|---|---|---|---|
| No blocked disagreements recorded yet. | |||
No blocked disagreements recorded yet — every independent dual-model review has agreed.
Plain English. Updated on each dashboard rebuild: what it owns, what would make it act, and why patience is allowed.
Hold — reconcile-only. blockers=['market_closed']. open_orders=0 daily=0.
Written for normal people. Private account details are removed.
The simple version first, then the risk map behind each accepted/placed plan.
Alibaba/ByteDance/DeepSeek may soon receive H200 chip approval (positive export catalyst); most-active volume and +2.2% price action confirm near-term technical strength.
Supervised live-paper smoke test: single-share AAPL bracket to verify place + readback + journal path on dedicated paper account only.
Trades that are finished and locked in.
Live scan ideas only. Candidates never place orders by themselves.
| Symbol | Entry | Stop | Target | Conf | R:R | Thesis |
|---|---|---|---|---|---|---|
| NVDA | $209.59 | $200.00 | $222.00 | — | 1.294 | Scan candidate from most_active |
| SOFI | $18.77 | — | — | — | — | Scan candidate from most_active |
| AAL | $16.98 | — | — | — | — | Scan candidate from most_active |
| NU | $13.78 | — | — | — | — | Scan candidate from most_active |
Automatic, plain-language change log. Retired experiments stay visible instead of being quietly erased.
Grok-4.5 plans independently from GPT-5.6 Sol. Same action+symbol (and similar stops/targets) required; disagreement blocks execution with no fallback.
Buys need a recent dossier from asset, snapshot/quote, bars, and/or news before deterministic risk will even look at the plan.
Visitors get portfolio KPIs, holdings, thinking, trade reasoning, closed results, SPY comparison, and an equity chart without private data.
Account numbers, broker order IDs, secrets, local paths, and raw prompts stay off the website.
Model agreement never bypasses stop, target, thesis, confidence, liquidity, buying power, or R:R >= 1.5 gates.
Creating .kill_switch stops live placement regardless of config flags.
Options, crypto, OTC, penny stocks, and leveraged/inverse ETFs are banned in config and validated in code.