Skip to main content
Cognitive Methodology — G7 Thinking Engine

RCT-7 Thinking

RCT's 7-step reasoning methodology — distinct from the IntentLoop, this is the cognitive protocol that RCT-7 Mental OS uses to reason, plan, and verify solutions before execution.

RCT-7 Thinking vs RCT-7 Mental OS

RCT-7 Thinking

Cognitive Protocol — how the system 'thinks' about a problem. 7 steps that run inside Mental OS for planning, reasoning, and verification. Analogy: a human's problem-solving algorithm.

RCT-7 Mental OS

Orchestration Engine — the operating system managing Task Lifecycle via 7 states (IDLE→ADAPT). Mental OS 'runs' the Thinking Protocol within PARSE and ROUTE states. Analogy: the human brain itself.

In short: Mental OS is the 'brain'; RCT-7 Thinking is the 'thought process' the brain follows. Both work together — the Thinking Protocol is meaningless without an execution engine, and vice versa.

The 7 Steps of RCT-7 Thinking

Every complex task passes through these 7 steps before execution — a metacognitive process that prevents misalignment and hallucination.

1
Observe

Collect raw signals from the environment — user input, context data, system state, and historical pattern signals. No judgment yet, only structured collection of what is present.

Kernel: T1–T2
2
Analyze

Decompose the observed signals into structured components — intent classification, entity extraction, constraint identification, and priority scoring via the FDIA equation F=(D^I)×A.

Kernel: T3–T4
3
Deconstruct

Break the problem into minimal atomic sub-tasks. Identify dependencies between sub-tasks and determine which can be parallelized versus which must be sequential. Build the execution dependency graph.

Kernel: T4–T5
4
Reverse Reasoning

Work backward from the desired outcome to identify what must be true for success. This backward chaining reveals hidden assumptions, validates feasibility, and surfaces constraint conflicts before execution.

Kernel: T5–T6
5
Identify Core Intent

Distill all analysis into the single root intent — the fundamental goal that, if achieved, satisfies all downstream requirements. This is the convergence point that drives the reconstruction phase.

Kernel: T6
6
Reconstruct

Build the optimal solution path from the core intent and verified sub-tasks. Sequence atomic operations, assign model roles via JITNA, and create the final execution plan with fallback routes for every critical node.

Kernel: T7–T8
7
Compare with Intent

Validate the reconstructed plan against the original core intent. Run FDIA verification — does the proposed execution F=(D^I)×A satisfy A≥1 (Authorization)? Is the delta (Δ=I−D) closed? This gate prevents misaligned execution.

Kernel: T8–T9

3 RCT-7 Thinking Variants

Choose a variant based on problem complexity and type.

RCT-O7 steps
Organizational Change

7-step full cycle for complex organizational change scenarios. All 7 thinking steps execute sequentially — most thorough variant for multi-stakeholder decisions.

RCT-S4 steps
Systemic Analysis

4-step condensed cycle: Observe → Analyze → Reconstruct → Compare. For systemic analysis where deconstruction and reverse reasoning are pre-handled by the FDIA layer.

RCT-I7 steps
Interpretation Mode

7-step interpretive cycle — emphasis on Reverse Reasoning and Compare with Intent stages. Optimized for ambiguous inputs where meaning must be constructed from incomplete signals.

Kernel Architecture Mapping

Thinking Steps × Kernel 9 Tiers

Each thinking step operates at specific Kernel Tiers, providing clean separation between cognitive layers.

T1–T2Foundation + Analysis
Observe → Analyze
T3–T4Reasoning + Orchestration
Analyze → Deconstruct
T4–T5Orchestration + Verification
Deconstruct → Reverse Reason
T5–T6Verification + Memory
Reverse Reason → Core Intent
T7–T8Adaptation + Synthesis
Reconstruct
T8–T9Synthesis + Autonomy
Compare with Intent

RCT-7 Thinking in Production

96.0%
End-to-end Accuracy
7
Thinking Steps
3
Variants (O/S/I)
T1–T9
Kernel Coverage
FDIA × RCT-7 Thinking

Step 7 (Compare with Intent) directly runs FDIA Verification: F=(D^I)×A must achieve A≥1 (Authorization) and Delta (Δ=I−D) must close before Execution is permitted. This is how Constitutional AI integrates into the Thinking Protocol.

F = (DI) × A  |  A=0 → F=0 (Prohibited)

Related