Episode 15: Running Your First AI Council Deep Audit

Welcome to TheFinalTape Academy – Episode 15: Running Your First AI Council Deep Audit

The AI Council Deep Audit transforms months of disciplined journaling—completed trades, honest checklists, ruthless Playbook tags, excursion metrics, and canonical PnL/R—into a prioritized, quantified action plan. It is not a generic summary or motivational pep talk. It is an 8-specialist war council (Performance Analyst, Behavioral Psychologist, Execution Tactician, Risk Assassin, Setup Surgeon, Regime Cartographer, Entry & Exit Judge, Chief Coaching Officer) that debates your recent performance like prosecutors dissecting a case.

The output is actionable:

• Kill List — ranked leaks with estimated preventable-loss impact ($), severity, agent quotes, and specific fix recommendations
• Reality Check — letter grade (A–F), trajectory diagnosis (Accelerating / Plateau / Declining), biggest opportunity, and primary threat
• Situation Room — full transcript of the agents’ debate (direct, often savage)
• Setup DNA — forensic breakdown per setup (regime-specific performance, compliance patterns, tag clusters)
• Evolution — grade and leak trend over time (requires multiple historical runs)

None of this exists until you explicitly trigger the analysis.

Where to Find It (March 2026 UI)

• Left sidebar → AI Council (or Kill List skull icon, depending on release labeling)
• Page loads with header: “AI Council – Your personal trading analysis powered by 8 AI specialists”
• Top section: Council Overview KPI cards + Date Range selector
• Tabs below header:

1. AI Council Audit (default, skull icon) — Kill List view
2. Reality Check
3. The Situation Room
4. Setup DNA
5. Evolution

Prerequisites – The Four Non-Negotiables

Before the Run button becomes active, these must be satisfied:

| Requirement | Why It Blocks the Analysis | |------------------------------------|-----------------------------------------------------------------| | Logged in | Edge Functions require valid JWT session authentication | | Portfolio selected | Analysis scopes to the active portfolio (top-right switcher) | | Valid week-aligned date range | Backend pipeline requires explicit ISO week boundaries (Monday–Sunday) | | ≥1 completed trade in the period | No closed trades = no data = immediate polite rejection |

Step-by-Step: Triggering Your First Deep Audit

1. Select a Valid Analysis Period

Use the Date Range selector above the tabs:

• Recommended first run: Last 12 Weeks (default preset when page loads)
• Other presets: Last 4 Weeks, Last 8 Weeks, Last 6 Months, Last Year
• Custom Range: Manually select start and end dates (both required, must form complete weeks)

Label updates dynamically, e.g., “Last 12 Weeks (Jan 27 – Mar 15, 2026)” → If the range is invalid (e.g., “All Time,” partial weeks, or no end date), the Run button remains disabled. Fix the range.

1. Navigate to the AI Council Audit Tab

Click the AI Council Audit tab (skull icon, usually first). If no report exists for the selected period:

• Large info card displays: “No AI Council Audit for This Period” or “Select a valid analysis period…”
• Two buttons appear:
• Run Analysis (Last 12 Weeks) — primary trigger (label reflects current preset)
• Check Again — refetches without consuming quota (use first if suspecting another device already ran it)

1. Trigger the Analysis

Click Run Analysis. Button changes to “Running Analysis…” with spinner and becomes disabled. Cost: 1 weekly AI analysis quota (resets Monday UTC). Important: Do not close the tab or refresh during execution — you will lose real-time progress visibility.

1. Monitor Progress Messages

Sequential messages appear below the button (60–120 seconds typical, depending on trade count):

| Visible Message | What’s Happening Behind the Scenes | |------------------------------------------------------|-----------------------------------------------------------------| | Step 1: Checking session… | Supabase JWT validation / refresh | | Step 1: Computing metrics (Last 12 Weeks)… | `compute-agent-metrics` Edge Function processes completed trades | | Step 1 complete: Analyzed N trades. Running Deep Audit… | N = count of completed trades; 0 → error and halt | | Step 2: Running 8-agent Deep Audit (1–2 minutes)… | `trading-agents-2026` orchestrates the 8 specialists | | Analysis complete! Loading results… | Results written to database; frontend refetches tabs |

Progress bar advances slowly during the agent debate phase.

1. Success – What You See
   • AI Council Audit tab loads the Kill List: ranked issues, total estimated preventable leakage ($), expandable cards with severity, agent quotes, and fix suggestions.
   • Other tabs populate: Reality Check (grade + plan), Situation Room (debate transcript), Setup DNA, Evolution (if historical runs exist).
   • Header displays last run timestamp and “Run New Analysis” / “Refresh” buttons.

Common Failure Modes & Immediate Fixes (March 2026)

| Symptom / Error Message | Root Cause | Fix | |-------------------------------------------------------|-------------------------------------------------|----------------------------------------------------------------| | Run button disabled / grayed | Invalid or missing week range | Select a preset (e.g., Last 12 Weeks) or set both custom dates | | “No completed trades found for the selected week…” | Zero closed trades in period | Choose range with completed trades or close open positions | | “Session expired. Please refresh… (Ctrl+Shift+R)” | JWT token stale/invalid | Hard refresh page, re-login if needed, retry | | “User or portfolio not found” | No active portfolio selected | Select portfolio from top-right switcher | | Stuck on “Running…” >3 minutes | Edge Function timeout or network issue | Refresh page, use “Check Again,” or wait and retry | | No report appears after “Analysis complete!” | Fetch race condition | Click Refresh button or reload page |

First-Run Pro Recommendations

• Start with Last 12 Weeks (default) — balances recency and sample size.
• Ensure ≥20–30 completed trades in the period — minimum viable signal for meaningful agent debate.
• After first run: Read Kill List top-to-bottom → select one item with highest estimated $ impact → implement fix next week.
• Do not attempt to fix everything at once. One deliberate change per audit cycle compounds faster and avoids overwhelm.

The AI Council does not flatter. It reads your canonical PnL/R, compliance scores, Playbook tags, excursion metrics, exit patterns, and regime context — then delivers unvarnished truth about where you are bleeding money and exactly how to stop it.

Run it. Read it without excuses. Act on one item. Or keep pretending the leaks are “just market conditions.”

The agents are waiting. They have read your journal. They know exactly where you are sabotaging yourself.

Next Episode: Reports & Analytics Deep Dive – Drilling from the 8 canonical metrics, Trade History, and AI Council outputs into regime-specific performance, taxonomy heatmaps, compliance violation clusters, exit quality distributions, drawdown forensics, and the complete autopsy suite.

Proceed once you have successfully run your first Deep Audit and reviewed the Kill List. One prioritized fix. One week. One compounding step. The council has spoken. Now listen — and execute.