We cannot evolve a system based on "raw text." We must parse thoughts into architectural primitives. This guide explains the mechanism of translating human intent into system structure.
The interactive playground has been moved to a dedicated full-screen tool for better usability and analysis history.
Every thought stream, no matter how chaotic, contains four fundamental signals that the system must extract. We group Ambiguity under Tension because uncertainty is a form of cognitive friction.
The Fuel. A signal of friction or gap between the Mental Model and Reality.
Types:
The Vector. A directional desire for a future state.
Input State: "I want..."
Output State: Proposition ("We should...")
Intent is the motivation; Proposition is the formalized claim.
The Structure. A bounded definition of a "thing" (Noun).
The Anchor. A rule or truth that must remain stable.
Let's analyze a real thought from the stream to see how the parser breaks it down. This example is backed by a typed data structure defined in thought-analysis-types.ts.
ComponentSubsystemIf we treat Component as Subsystem, it inherits System Theory hierarchy
Components gain Sensors and Levers for observability and control
Gall's Law / Systems Theory
cf. John Gall, Systemantics (1975)
Fails if a trivial component (e.g., a Button) requires no internal state or control
The output of parsing a thought is not a "Ticket" (work to be done). It is a Design Hypothesis (a testable proposition about system structure). This distinction is critical for evolutionary architecture.
An atomic unit of potential system evolution. It is not yet a fact; it must be validated.
Input stream. Chaotic, unstructured, high-volume.
You are here. A parsed, testable proposition.
Prototyping, debate, or simulation to test the hypothesis.
Accepted structure. Becomes an Architectural Decision Record.
The basic parser captures the Logic, but logic is not enough for a living system. To fully understand a thought, we need to capture the context and emotion.
Why it matters: Thoughts don't exist in a vacuum. Knowing what triggered this thought (e.g., "Frustration with current code") changes how we prioritize it.
Why it matters: "wowowowow!!!" is a high-confidence signal. A dry statement is low-confidence. We prioritize high-energy insights.
Why it matters: We must ask "What would prove this wrong?". Without this, we build on fragile assumptions.
The framework is not just conceptual—it is backed by a precise TypeScript type system. This enables machine parsing, validation, and traceability.
// Core Primitives
type TensionType = 'Semantic' | 'Structural' | 'Logical';
interface Tension {
type: TensionType;
description: string;
conflictingTerms?: [string, string];
}
interface Intent {
rawExpression: string;
targetDomain?: string;
}
interface Proposition {
claim: string; // "If we structure X as Y..."
expectedOutcome: string; // "...then Z follows."
derivedFromIntent: Intent; // Traceability
}
// The Design Hypothesis
interface DesignHypothesis {
id: string;
proposition: Proposition;
predictions: string[];
falsifiability: string;
theoreticalBasis?: { source: string; citation?: string };
tensions: Tension[];
status: 'Draft' | 'UnderReview' | 'Validated' | 'Invalidated';
}More examples of applying the framework to real thoughts. These demonstrate how Conceptual Tensions translate into concrete System Changes.
PullDemandIf we introduce a "Pull Mechanism" that activates on Component creation/evolution, then the system becomes self-organizing
Creating a Component will automatically generate requirements; modifying a Component will identify and queue affected artifacts for realignment.
Systems Theory / Demand-Pull Systems
cf. Lean Manufacturing (Taiichi Ohno): Pull systems vs. Push systems.
Fails if a Component can be created without any downstream work. Fails if Component modifications do not require realignment of any artifact.
PersuasionAuthenticityIf we build an automated Ugly Website Pipeline (Crawler → AI Redesigner → Pitch Generator), then we can generate qualified leads with minimal effort
5 minutes of work produces a €500 quote opportunity
Lean Startup / Cold Outreach
cf. Eric Ries (Lean Startup), Aaron Ross (Predictable Revenue)
Fails if: (1) Effort per lead exceeds 15 minutes, (2) Conversion rate is <1%, (3) Business owners perceive outreach as spam, or (4) AI redesigns are not compelling.
EpisodesActs/Scenes/PerformancesActorsSingers/PerformersIf we build a structured Opera Discovery App with entities for Operas, Venues, Performers, and Productions, then users (enthusiasts and newcomers) can explore the art form
Users will engage with opera content, discover new works, and develop appreciation for the art form
Content Discovery / Cultural Tech
cf. Spotify (music discovery), IMDB (film database)
Fails if: (1) Users bounce within 30 seconds, (2) Enthusiasts find content too shallow, (3) Newcomers find content too jargon-heavy, or (4) Data model cannot represent opera complexity.