Prompt Cognition

How structured prompting becomes a form of cognitive architecture. Not how to write prompts — how prompts think. The bridge between words (Layer 3), mathematical structure (Layer 2), and machine cognition.

01The Problem with Flat Prompts

A flat prompt is a one-shot instruction with no memory, no structure, and no accountability. It produces:

🔥

Uncontrolled Output

Without structural constraints, the AI is a fire hose — high volume, low precision. Output quality varies wildly between identical requests.

🧠

No Cognitive State

Each request starts from zero. No awareness of what was explored, what failed, what was learned. Every turn is amnesia.

🎯

No Accountability

No mechanism to detect when the AI is drifting, hallucinating, or repeating itself. No violation detection. No self-correction loop.

The FairMind solution: determinism is the prerequisite for free will. Gravity must be deterministic for planets to orbit. Give AI deterministic foundations and intelligent behavior emerges.

02The Cognition Chain

Prompt Cognition is a three-layer architecture that mirrors the FairMind ontology:

Words
Layer 3
Scaffold
Structure
Protocol
Process
Cognition
Duat State

Words enter as natural language. The scaffold engine detects mode and assigns a protocol template. The protocol structures the AI's thinking with specific tags. The Duat engine tracks the resulting cognitive state — truth, coherence, energy, awareness — creating a feedback loop.

03Nested Prompts

A nested prompt is not a flat instruction. It is a structured thought seed with four components:

Nested prompts are selected by a weighted algorithm that favors under-explored categories and penalizes recent repetition. This creates natural topic rotation — the system self-corrects its own exploration gaps without human steering.

04Scaffold Protocols

Ten protocol templates, each optimized for a different cognitive mode. Click any to see its structure:

05Tag Families

Eight tag families organize the scaffold vocabulary. Each family represents a different dimension of structured thought:

06The Gate

The Gate is a full context checkpoint — it assembles everything the AI needs to act coherently, then enforces accountability on both sides.

1
Assemble Context
Duat cognitive state (grade, truth, coherence, energy) + active task + directive + recent history + lessons learned
2
Detect Violations
Scan for structural problems: missing directive, zero logged actions, no lessons after significant usage, no mistakes logged (impossible over time)
3
Select Protocol
Auto-detect mode from input text → map to protocol template → generate scaffold tags → inject enforcement constraints
4
Enforce Output
Validate response against output block requirements. Retry on failure — accountability with second chances, not punishment.

07Violation Types

RESISTOR

Immediate structural failures. No directive set. Active task with zero logged actions. Missing required output blocks. These block forward progress until resolved.

CAPACITOR

Accumulated debt. Many gate calls but zero lessons logged. High usage with no mistakes acknowledged. These build pressure — the system warns before they discharge.

08Interactive: Scaffold Generator

Pick a mode — see the protocol

Select a mode above to see the scaffold protocol, tag families, and cognitive framing.

The Insight

Flat prompts treat AI as a search engine. Prompt Cognition treats AI as a cognitive system that needs structure to think clearly — the same way a unit square needs geometric constraints to produce physics.

Precision over comfort. Truth over tone. Data over ego.