Night Hawk Cortex: The Evidence Brain Behind Every Commit

Most trading systems reduce a dozen indicators to a single score and act on a threshold. The Night Hawk Cortex does something fundamentally different: it composes evidence from eight independent data sources, each timestamped and signed, into a structured verdict where any single source can veto an entry but no single source can force one. This is a precision-first architecture — designed to keep you out of bad trades rather than to get you into more trades.

The eight evidence sources

Each source reads a distinct slice of the market. They are composed in a fixed order for deterministic output, but ordering never affects weighting:

1. GEX walls — Reads the Vector GEX ladder: is the play's target path blocked by a dominant opposing dealer wall? Checks wall positioning, regime-style match (long gamma vs. short gamma), and path clearance. Veto-capable: a dominant wall sitting directly in your path vetoes the commit.

1. Wall lifecycle — The flagship wall-trend signal. Reads the session wall-history rail (the beads' lifecycle over time) to determine whether walls are building, holding, or dissolving. A wall that has held all session is a materially different structure from one that appeared two minutes ago.

1. Flow quality — Reads the HELIX print texture: sweep clusters, block trades, opposing whale activity. Veto-capable: an opposing sweep/block cluster exceeding $1M within 15 minutes vetoes the commit — an institutional-size bet against your thesis.

1. Sector heat — Reads Thermal sector alignment and market breadth tone. A bullish tech setup is stronger when the technology sector leads; a bearish one is stronger when breadth is negative.

1. Catalyst/news — Reads Benzinga news channels through deterministic tagging (headline keywords, channel classification). No LLM in the money path — the discrimination is rule-based and reproducible. Earnings-day prints get special treatment: an after-hours report into a long-premium position opposes.

1. VEX/charm — Reads the net dealer dollar-vanna across the GEX matrix and the "king" node (the strike with the most net gamma). Models charm as a time-of-day heuristic against pin distance — how theta decay at different points in the session affects the dealer hedging anchor.

1. Dark-pool confluence — Reads institutional dark pool levels and checks whether they cluster near gamma walls. When dark-pool institutional accumulation and dealer positioning point to the same zone, the confluence bonus fires.

1. Opening harvest — A time-locked window (9:30-9:45 ET) that reads the character of the open: the overnight gap versus the prior close, the opening drive's direction and conviction, and market internals. This source self-silences after the first 15 minutes.

Evidence decay: stale data self-silences

Every evidence item carries an exponential half-life measured in seconds. Fresh evidence contributes its full weight. As time passes, the contribution decays exponentially. After 3 half-lives (at which point the item is worth less than 12.5% of its original weight), the evidence is demoted to absent — the honest statement "this source cannot answer right now" rather than a microscopic weight pretending to be a real signal.

This decay is critical for 0DTE. A wall that was dominant at 10:00 AM may not be dominant at 1:00 PM. A flow cluster from 9:35 means less by noon. The Cortex does not carry stale readings as if they are fresh — it explicitly ages them out and marks them absent when they have expired.

Veto asymmetry: the precision-first design

The most important architectural decision in the Cortex is its asymmetry between supports and vetoes:

- Supports are capped per source. No single bullish signal — no matter how loud — can by itself force a commit. Each source has a documented support cap, and the total supporting score is the sum of these bounded contributions. - Vetoes are unbounded hard blocks. A single veto from any veto-capable source (GEX walls or flow quality) blocks the commit regardless of the total supporting score. A dominant opposing dealer wall in the play's path kills the entry even if every other source supports it.

This means the Cortex cannot be "out-scored." A 93-score setup with a wall in its path is blocked. A setup with five sources supporting but one $1.5M opposing sweep cluster is blocked. One loud bearish fact can kill an entry; one loud bullish signal can never buy one.

When both veto-capable sources fail to read entirely (provider outage, timeout), the system enters a veto-blind state and holds fresh commits rather than letting them through without dealer-wall and whale-flow protection. Opening new risk blind to both channels is exactly what fail-closed is for.

Conviction bands

The composed score maps to conviction:

- A (score >= 2.0) — requires the dealer landscape AND its lifecycle to both argue for the play, or equivalent breadth across multiple smaller sources. - B (score >= 0.75) — one full mid-tier signal net of opposition. A real edge, but not a structural argument. - C (below 0.75) — nothing here earns size.

Display conviction is capped at A. The A+ tier is reserved for the tier system's record-based unlock — 10+ plays at 80%+ win rate — not for the Cortex score alone.

The full vector is pinned at commit

When a play commits to the ledger, the complete evidence vector is frozen and pinned to the ledger row. This is a first-write-wins model: the evidence that justified entry is preserved exactly as it was at commit time, even as the live sources continue to update.

This pinned snapshot serves three purposes: it lets the thesis health system compare "what I believed at entry" to "what the evidence says now," it lets the calibration loop measure whether Cortex scores predicted outcomes, and it gives the member an honest audit trail of why the trade was taken.

Exit mechanisms driven by evidence

The Cortex does not only gate entry. Two exit mechanisms flow from the evidence:

- Thesis-break exit — when the evidence that justified entry has not merely faded but actively reversed (health rung: BROKEN or OPPOSITE), the system triggers an immediate unconditional exit. This is not a stop-loss; it is an evidence-driven "the reason for this trade no longer exists." - Profit ratchet — the floor only ratchets up, never down. As the position moves in your favor, the minimum acceptable exit price rises. This interacts with the thesis health — a trade where the thesis is intact holds its runner; one where the thesis is weakening locks in more aggressively.

How this differs from indicator systems

Traditional systems stack indicators (RSI, MACD, moving-average crossovers) and act on a threshold. The Cortex differs in three structural ways: (1) every evidence item is signed and timestamped, not a snapshot-in-time number — it decays and self-silences; (2) the veto asymmetry means the system is precision-first, not recall-first; and (3) the full evidence vector is pinned and auditable on every committed trade, so performance can be traced to specific evidence patterns rather than aggregate scores.

For the gate stack that runs before the Cortex, see the 0DTE Command guide. For the evening playbook the Cortex also feeds, see the Evening Edition guide. For the underlying gamma exposure, options flow, and dark pool data the sources read, see those deep-dives. Get access →

BlackOut provides educational tools and market analysis only and does not provide investment advice. Options and equities trading involve substantial risk and are not suitable for every investor.