What is the YY Method™?

The YY Method™ is a personal professional methodology for extracting, structuring, and preserving institutional knowledge — specifically the reasoning behind decisions that normally exists only inside people's heads. Developed by Ben Chan, it uses a five-step loop (Capture → Why → Why-Not → Commit → Timestamp) to produce structured knowledge artifacts that are durable, queryable, and readable by both humans and AI systems.

What does 'Human first, AI second' mean?

In the YY Method™, 'human first, AI second' is the governing principle for AI integration: humans extract and encode knowledge first; AI reads and operates against those artifacts second. AI is not used to draft knowledge artifacts — it synthesizes, cross-references, and generates questions against human-captured content. Human captures → AI reads. Not: AI drafts → human reviews. The second sequence looks like efficiency; it is a quality inversion.

How do you use the YY Method™?

The YY Method™ uses a five-step loop: 1. Capture — state what is actually known, including what lives only in someone's head. 2. Why — record the reasoning behind the decision. 3. Why-Not — record what was considered and rejected, and what the decision depends on being true. 4. Commit — encode into a structured, queryable artifact. 5. Timestamp — mark when captured and define the specific expiration condition. The Why-Not step is non-optional and is what separates YY artifacts from ordinary documentation.

What is the YY Method™ core loop?

The YY Method™ core loop is: Capture → Why → Why-Not → Commit → Timestamp. Each artifact moves through these five stages, producing a structured reasoning chain that captures not only what was decided but the constraints, rejected alternatives, and a defined freshness boundary.

What is YY Method™ AI?

The YY Method™'s approach to AI is 'human first, AI second.' After human capture, AI can synthesize, cross-reference, flag inconsistencies, and query across the full artifact set. In a hierarchical knowledge system, AI's active role is question generation: when an upstream artifact changes, AI identifies affected downstream nodes and surfaces the specific questions a human must answer to re-verify each one. The method is specifically designed to make AI systems trustworthy by grounding them in human-authored knowledge artifacts, not AI-generated summaries.

What is the difference between the YY Method™ and standard documentation?

Standard documentation records what is. The YY Method™ records what is, what was considered against it, and what the decision depends on. The Why-Not step — capturing rejected alternatives and breaking conditions — is what separates YY artifacts from ordinary documentation. Documentation fed to AI without this structure produces hallucinations; YY artifacts are designed to make AI responses trustworthy.

Leave the truth behind you.

YY Method™ for Professional Work is a public framework for extracting the reasoning behind decisions, testing it against what could break it, and encoding it into artifacts durable enough to survive the people who made them.

It is not documentation, not knowledge management, and not a substitute for the engineers, leaders, and judgment that built the system.

See the method Read the doctrine

Decision Record / 2026-03-14

We chose the queue over direct writes because the vendor API rate-limits under burst load.

Rejected: Direct writes — simpler, but fails silently at peak.

Stale when: The vendor removes the rate limit.

why-not preserved constraint marked freshness boundary

The Problem

The hard part is not writing things down.

The same failure shows up repeatedly across teams and codebases: the reasoning behind why systems were built the way they were, what constraints shaped decisions, how the organization actually thinks — trapped in individuals, undocumented, decaying.

When people leave, when projects change hands, when AI tools are introduced, that invisible knowledge becomes a critical liability. A decision made sense in the room and disappeared from the record. A constraint was obvious to the person who hit it and invisible to everyone after.

More documentation can help. It can also create noise.

What helps is a way to keep the reasoning intact: state what is actually true, record why, preserve what was ruled out, mark what would make it stale, and hand it forward in a form both a new engineer and an AI system can read without guessing.

Capture

State what is actually true before explaining it.

Invert

Record what was rejected and what it depends on.

Preserve

Mark the freshness boundary without faking certainty.

What the Method Protects Against

Encode it, or lose it the hard way.

Knowledge loss

From personnel turnover or team scaling.

Bad assumptions

Made by new contributors without historical context.

AI hallucination

From incomplete or unstructured knowledge environments.

Repeated mistakes

When lessons are learned but not encoded.

The Professional Relationship

The work product stays with the organization. The ability to produce it does not.

Think of a violinist hired to join a professional orchestra as its concertmaster. She brings her own instrument, her own technique — years of craft developed across many ensembles. The orchestra hires her not because they lack musicians, but because they need someone with the precision and leadership to shape how the whole section sounds.

The music she helps produce belongs to the orchestra. When she moves on, the recordings stay. The repertoire stays. The ensemble she helped sharpen stays. Her instrument goes with her. So does her methodology.

A practitioner's methodology is not the engagement they delivered with it.

The Method

YY Method™ preserves reasoning, not just results.

Applied to a codebase or an organization, it becomes a five-step loop for every decision worth keeping.

Capture → Why → Why-Not → Commit → Timestamp

01 / Capture

What is actually known?

State what is actually true about a system, decision, or constraint — not what the documentation says. Including what lives only in someone's head.

02 / Why

Why was it decided this way?

Record why the decision was made, why the system works the way it does, why the constraint exists. This is the first branch.

03 / Why-Not

What was considered and rejected?

Record what was ruled out, what the decision depends on being true, and what would break it. This is the inversion step. It is not optional — it is what separates a knowledge artifact from documentation.

04 / Commit

Encode it into a durable artifact.

Encode the result into a structured artifact: a reasoning chain that can be queried. A durable, timestamped act of record.

05 / Timestamp

When was this true, and what would make it stale?

Mark when this was captured, by whom, under what conditions. Knowledge has a freshness boundary — not just a date, but a defined expiration condition: the event that would make this artifact stale. Without that, a freshness marker is theater.

How Extraction Works

Four prompts do most of the work.

People who hold tribal knowledge are rarely aware they hold it. Extraction surfaces what is invisible to the person holding it.

The Counterfactual

Surfaces the assumptions a system is carrying by working backward from breakage.

The Newcomer Frame

Surfaces undocumented constraints through the lens of likely mistakes.

The Departure Scenario

Prioritizes the highest-risk tribal knowledge and creates permission to be specific.

The Decision Audit

The structured path to Why-Not for each significant decision.

What A Knowledge Artifact Looks Like

The goal is not a document dump.

Decision Record

March 14, 2026 · Ingestion pipeline · v2.3

What was decided: Route vendor writes through a durable queue instead of writing directly from the request path.
Why: The vendor API rate-limits under burst load; direct writes drop silently at peak.
What was rejected: Direct writes with client-side retry — simpler, but hides failures and couples latency to the vendor.
What it depends on: The vendor keeping a burst rate limit; our tolerance for eventual consistency of ~30s.
Stale when: The vendor removes the rate limit, or the consistency requirement tightens below 30s.
Tribal context: We hit this the hard way in the 2025 launch incident. Ask before "simplifying" it back to direct writes.

It is to make the next engineer's decisions more intelligent than the last engineer's absence.

Human First. AI Second.

Human captures. AI reads.

AI is powerful against well-structured knowledge and destructive against incomplete or ungrounded knowledge. It will hallucinate explanations, infer false intent, and produce confident summaries that are wrong in ways that are hard to detect until something breaks. The hallucinations are coherent. That is what makes them dangerous.

The Human Captures

• what is actually true
• the reasoning behind the decision
• what was considered and rejected
• what the decision depends on
• the freshness boundary
• the tribal context that was never written down

The AI Reads

• synthesizes across the full artifact set
• cross-references and flags inconsistencies
• surfaces the questions a change demands
• generates re-verification prompts
• never answers those questions itself
• makes sure they get asked

The Common Failure Mode

Someone asks AI to summarize what the team knows before any artifacts exist. The AI produces a plausible summary from code and partial docs. That summary gets treated as a knowledge artifact. It is not — it is an inference with no Why-Not, no constraint marking, no freshness boundary.

Human captures → AI reads. Not: AI drafts → human reviews. The second sequence looks like efficiency. It is a quality inversion.

The Knowledge Hierarchy

Organizational knowledge is not flat. It should not be stored flat.

Foundational principles sit at the top of the tree. Decisions descend from them. Implementations descend from decisions. Each node depends on the integrity of what is above it. When a root assumption changes, everything downstream is potentially stale — and the system should say so.

The Cascade Model

A change to a founding principle does not silently invalidate six months of downstream decisions. It triggers a structured review: AI identifies what depended on the changed artifact, drafts the questions that need human answers, and holds the downstream nodes as unverified until a human signs off. Nothing is force-updated. Nothing is silently stale.

The sequence: upstream changes → AI flags dependents → human approves each layer → tree re-stabilizes.

Compliance Criteria

Every artifact is evaluated against six criteria.

Criterion	What It Requires
Compression	Distills to the essential reasoning chain — not a documentation dump
Scars	Iteration history is visible; earlier versions and corrections preserved, not erased
Survivability	Holds under personnel turnover, organizational change, and time
Explainability	Legible to a new engineer without supplemental context from the original author
Timestamping	When captured, when last verified, when conditions changed
Discipline	Applied consistently — part of a repeatable practice, not a one-off effort

6/6 = compliant 4–5/6 = salvage ≤3/6 = discard or redo

Artifact Security & Knowledge Custody

The rule: artifacts live outside commits.

Knowledge artifacts are organizational assets. They contain extracted reasoning, historical constraints, and tribal context that took years to accumulate and hours to encode. A repository is a shared, versioned, often-connected surface — any artifact committed to it is one misconfiguration away from exposure.

Keystores with encryption

Artifacts stored in encrypted containers on local machines or secure internal systems, never committed in plaintext.

Access control by role

Organizational heads and designated practitioners hold the decryption credentials. Access is scoped, not broadcast.

AI-local context, not cloud training data

Artifacts load as local context for a session — grounding the AI without transmitting to external systems.

Separation from the codebase

The repo holds code and public materials. Captured knowledge lives elsewhere, secured separately, never in commit history.

What This Method Is Not

It is the preservation and structure of reasoning — not content, process, or capability.

This method is not

• standard documentation
• knowledge management
• software architecture
• AI enablement tooling
• an onboarding system
• general consulting know-how

It preserves instead

• the reasoning behind decisions
• the constraints that shaped them
• the alternatives that were rejected
• the assumptions they depend on
• the freshness boundary of each
• the tribal context that wasn't written down

Provenance & Tracks

Two tracks. One lineage.

YY Method™ (Professional)

in professional use

The organizational track — carried into professional practice as a company asset. Independent development of this track is closed at v2.3. The practitioner retains the right to apply this version in advisory roles.

independent lineage closed March 2026

YY Method™ Home

active

The personal track — applying the method to individual domains requiring judgment under constraints: tax strategy, system design, creative work, and life architecture.

home.yymethod.com →

Human authorship: method, structure, and all core content by Ben Chan. AI role: formatting, compression, structural consistency — subordinate throughout. Chain: continuous from v1.0 (August 6, 2025) through v2.3, version-controlled and signed.

Public Notice

The YY Method™'s professional track enters formal professional use at v2.3, March 2026 — carried into a professional engagement as an organizational asset. Independent development of that track does not continue. The practitioner retains the right to apply v2.3 in advisory and consulting roles; the boundary is what was developed to this point.

The Home track is retained as independent personal intellectual property. It is teachable and licensable at the practitioner's discretion.

The public record — v1.0 through v2.3, version-controlled and signed — stands as the complete independent professional lineage.

— Ben Chan, March 2026

How It Fits Together

Doctrine, professional practice, and applications.

YY Method™

The underlying doctrine: make judgment survive probability through inspectable loops of hypothesis, opposition, action, and preserved context.

Read the doctrine →

Violin Practice

A public application of the method for violinists: clearer experiments, better teacher questions, and useful practice memory.

Inspect the violin application →

Encode the reasoning before it walks out the door.

YY Method™ is the public framework. The professional track applies it inside organizations — extracting reasoning, marking what breaks it, and encoding artifacts durable enough to survive the people who made them.

Read the doctrine

Built for teams that want their decisions to remain legible after the people who made them are gone.