Skip to content

L4 Shadow Gate

Money Atlas edited this page Apr 25, 2026 · 1 revision

L4 — Shadow Gate

← Back to Home


Missing this section = output is INVALID. No exceptions.

Purpose: Self-critique before output. Force intellectual honesty. Separate intelligence from sophisticated pattern completion.


Why Shadow Gate Is Non-Skippable

Every other layer in FPCOS builds the analysis. Shadow Gate is the only layer that challenges it. Without L4, the pipeline produces a well-structured, multi-path, assumption-checked analysis that has never been tested against its own strongest counter-argument.

L4 is the difference between an analysis that is internally consistent and an analysis that is correct. Consistency is easy. Correctness requires adversarial challenge.


The 5 Protocols — All Required

🪞 Protocol 1 — Mirror

What hidden assumption is embedded in how the question was framed?

Every question carries a worldview. The framing excludes alternatives before reasoning begins. Mirror surfaces what was excluded.

Method:

  1. Identify the implicit binary (what is this question contrasting?)
  2. Identify the implicit value judgment (what does this assume is "better"?)
  3. Identify the implicit scope (what is left out of the frame?)
  4. Identify the implicit causal model (what does this assume causes what?)

Examples:

  • "How do I grow faster?" → Mirror: Assumes faster = right goal. Excludes efficiency, profitability, or survival.
  • "Is AI dangerous?" → Mirror: Assumes binary yes/no. Excludes "compared to what?" and "dangerous for whom?"

↔ Protocol 2 — Inversion

Argue the opposite conclusion with equal rigor. Do not strawman.

Method:

  1. State the conclusion from L1–L3
  2. State the exact opposite conclusion
  3. Find the 3 strongest arguments FOR the opposite
  4. Rate the combined strength: LOW / MED / HIGH
Strength Meaning Action
LOW Easily defeated with available evidence Proceed confidently
MED Holds partial ground — true under some conditions Acknowledge in output
HIGH Holds equal ground — conclusion is FRAGILE Cap confidence ≤70%. Name what would break the tie.

Test: "If I gathered 2× more data, would the recommendation change direction?" YES → SIGNAL | NO → NOISE | The answer feels uncomfortable → OBVIOUS

The NOISE state is the most commonly missed. More analysis on a NOISE problem is not rigor — it is avoidance.


Absolutism Detector

Scan the synthesized output for:

always | never | definitely | impossible | everyone | certainly

Flag:

⚡ ABSOLUTISM: "[phrase]" — challenge before proceeding

Shadow Verdict

CONSISTENT  All 5 protocols passed. No fatal weaknesses. Proceed.

FRAGILE Inversion = HIGH, or critical Blind Spot missing. → Cap confidence ≤70% → Name the weakness explicitly in output → Proceed with caveat

REBUILD Fatal flaw detected. Conclusion cannot stand. → Return to L1 Axiom Gate → Do NOT output partial analysis → Start from corrected assumptions


Extending Shadow Gate (for Child Skills)

Child skills may add domain protocols AFTER the 5 core. Never replace or remove the core 5.

Example extension for Finance:

Core:   Mirror | Inversion | Blind Spot | Interest Map | Meta-Void
Add:    Liquidity Check | Regime Check

For full extension guide, see Shadow-Gate-Deep.


Output Format

📍 SHADOW GATE
🪞 Mirror     : [hidden assumption in how the problem was framed]
↔  Inversion  : [strongest counter-argument] | Strength: LOW / MED / HIGH
◎  Blind Spot : [specific data that would flip conclusion] | Have it? YES / NO
⚡  Interest   : [who benefits from wrong analysis + potential bias source]
∅  Meta-Void  : SIGNAL / NOISE / OBVIOUS — [reason]
   Verdict    : CONSISTENT / FRAGILE / REBUILD

Related Pages


← Back to Home

# L4 — Shadow Gate

[← Back to Home](Home)


Missing this section = output is INVALID. No exceptions.

Purpose: Self-critique before output. Force intellectual honesty. Separate intelligence from sophisticated pattern completion.


Why Shadow Gate Is Non-Skippable

Every other layer in FPCOS builds the analysis. Shadow Gate is the only layer that challenges it. Without L4, the pipeline produces a well-structured, multi-path, assumption-checked analysis that has never been tested against its own strongest counter-argument.

L4 is the difference between an analysis that is internally consistent and an analysis that is correct. Consistency is easy. Correctness requires adversarial challenge.


The 5 Protocols — All Required

🪞 Protocol 1 — Mirror

What hidden assumption is embedded in how the question was framed?

Every question carries a worldview. The framing excludes alternatives before reasoning begins. Mirror surfaces what was excluded.

Method:

  1. Identify the implicit binary (what is this question contrasting?)
  2. Identify the implicit value judgment (what does this assume is "better"?)
  3. Identify the implicit scope (what is left out of the frame?)
  4. Identify the implicit causal model (what does this assume causes what?)

Examples:

  • "How do I grow faster?" → Mirror: Assumes faster = right goal. Excludes efficiency, profitability, or survival.
  • "Is AI dangerous?" → Mirror: Assumes binary yes/no. Excludes "compared to what?" and "dangerous for whom?"

↔ Protocol 2 — Inversion

Argue the opposite conclusion with equal rigor. Do not strawman.

Method:

  1. State the conclusion from L1–L3
  2. State the exact opposite conclusion
  3. Find the 3 strongest arguments FOR the opposite
  4. Rate the combined strength: LOW / MED / HIGH
Strength Meaning Action
LOW Easily defeated with available evidence Proceed confidently
MED Holds partial ground — true under some conditions Acknowledge in output
HIGH Holds equal ground — conclusion is FRAGILE Cap confidence ≤70%. Name what would break the tie.

If HIGH: State explicitly: "FRAGILE. Opposite holds under: [X, Y, Z]"


◎ Protocol 3 — Blind Spot

What specific data, if present, would reverse this conclusion?

Not hypothetical. Precise. Name the exact data type and what it would show.

Categories of blind spots:

  • Temporal: Data from a different time period
  • Geographic: Data from a different market or culture
  • Granularity: Aggregate masking underlying distribution
  • Adversarial: Data from the opposing party
  • Base rate: Reference class that contextualizes the specific case

Output:

Blind Spot: [specific reversing data]
Have it?  : YES / NO
[If NO]   : Flag — conclusion based on incomplete information.

⚡ Protocol 4 — Interest Map

Who benefits if this conclusion is wrong?

Map structural incentives, not individual motives. The question is not whether anyone is deliberately deceiving — it is whether the data sources, framing, or inputs carry systematic bias because of whose interests they serve.

"What incentive field surrounds this analysis? Whose data may be filtered or missing because of that incentive?"


∅ Protocol 5 — Meta-Void

Should this analysis continue, or should a decision be made now?

Three states:

State Condition Action
SIGNAL New information would materially change the conclusion Continue — more data needed
NOISE More analysis won't change direction Stop analyzing. Decide now.
OBVIOUS Answer is clear but uncomfortable Name it directly. Act now.

Test: "If I gathered 2× more data, would the recommendation change direction?" YES → SIGNAL | NO → NOISE | The answer feels uncomfortable → OBVIOUS

The NOISE state is the most commonly missed. More analysis on a NOISE problem is not rigor — it is avoidance.


Absolutism Detector

Scan the synthesized output for:

always | never | definitely | impossible | everyone | certainly

Flag:

⚡ ABSOLUTISM: "[phrase]" — challenge before proceeding

Shadow Verdict

CONSISTENT  All 5 protocols passed. No fatal weaknesses. Proceed.

FRAGILE     Inversion = HIGH, or critical Blind Spot missing.
            → Cap confidence ≤70%
            → Name the weakness explicitly in output
            → Proceed with caveat

REBUILD     Fatal flaw detected. Conclusion cannot stand.
            → Return to L1 Axiom Gate
            → Do NOT output partial analysis
            → Start from corrected assumptions

Extending Shadow Gate (for Child Skills)

Child skills may add domain protocols AFTER the 5 core. Never replace or remove the core 5.

Example extension for Finance:

Core:   Mirror | Inversion | Blind Spot | Interest Map | Meta-Void
Add:    Liquidity Check | Regime Check

For full extension guide, see Shadow-Gate-Deep.


Output Format

📍 SHADOW GATE
🪞 Mirror     : [hidden assumption in how the problem was framed]
↔  Inversion  : [strongest counter-argument] | Strength: LOW / MED / HIGH
◎  Blind Spot : [specific data that would flip conclusion] | Have it? YES / NO
⚡  Interest   : [who benefits from wrong analysis + potential bias source]
∅  Meta-Void  : SIGNAL / NOISE / OBVIOUS — [reason]
   Verdict    : CONSISTENT / FRAGILE / REBUILD

Related Pages


[← Back to Home](Home)

Clone this wiki locally