PowerUp Coaching — Living Knowledge Base

SKILL

14 min read

Eigenthinking

Transform how you think into what you own.

In linear algebra, eigenvectors are the natural transformation axes of a system — the directions that produce maximum movement per unit of input. Every expert thinker has cognitive eigenvectors: the patterns of inquiry that consistently produce disproportionate insight. Most experts never name them. Eigenthinking does.

The core equation: UP → IP
Your Unique Process, made explicit, becomes Intellectual Property.

How to Run This Skill

Work through the nine steps sequentially. Each step has a clear completion signal — don’t move forward until that signal is present. The steps build on each other; rushing a step creates compounding problems downstream.

Use AI as a thinking partner throughout, not an answer machine. The user’s domain expertise is the source. AI is the capture mechanism, cross-domain pattern library, and production system.

For each step: read the step, run the practice with the user, verify the completion signal, then advance.

Reference files (load as needed):

  • references/nine-steps.md — Full step-by-step protocol with prompts and examples
  • references/up-to-ip-framework.md — The generalized SOP this skill instantiates
  • references/cognitive-patterns.md — Common expert cognitive patterns and how to surface them
  • references/production-pipeline.md — Multi-role artifact production (AIMM integration)
  • references/worked-example.md — Complete end-to-end case study: the Latent Cartographer session traced through all nine steps. Load this to calibrate what each step actually produces in practice — what a real SIGNAL sounds like, what drilling looks like in sequence, what stress-testing finds, what an organic insight moment looks like. The single best reference for quality calibration across the full cycle.

The Nine Steps (Overview)

SIGNAL → DRILL → REFRAME → NAME → BUILD → STRESS TEST → COMPRESS → LIFT → TEACH
StepIntentOutput
1. SIGNALFind structural friction worth followingOne-sentence friction statement
2. DRILLDiagnose the mechanism, not the symptomStructural explanation that satisfies a skeptic
3. REFRAMEBuild the new lens that makes solutions visibleOne-sentence reframe
4. NAMEMake the concept portable and ownableA name that creates a curiosity gap
5. BUILDTurn the reframe into a followable systemComplete named methodology with quality gates
6. STRESS TESTFind the honesty gap before the market doesHardened methodology with guardrails added
7. COMPRESSProve the thinking survives reductionOne-sentence + one-paragraph versions
8. LIFTMove one abstraction level upThe principle the methodology instantiates
9. TEACHProduce the artifact that transfers itA finished artifact that demonstrates, not just describes

Load references/nine-steps.md for the full protocol on each step.

Running the Skill: Conversation Protocol

Opening Move

Before starting the protocol, check for prior work — then assess where the user’s thinking already is. Not everyone arrives at Step 1.

Step 0 — Check for existing log:

At the start of every session, search the working directory for files matching eigenthinking-log-*.md. This is the first action — before asking the user anything.

  • If one log is found: Read it. This is the authoritative session record. Confirm the goal and current step with the user, then go directly to the Entry-Point Assessment below. Skip the Triage Question — the log replaces it.
  • If multiple logs are found: List them and ask which engagement the user wants to continue (or whether this is a new one).
  • If no log is found: This is a new engagement. Proceed with the Triage Question and Session Goal below. Create the log after the Session Goal is established.

The Triage Question (new engagements only — skip if a log was found):

“Before we start — have you already been working on this? Do you have a written problem statement, prior sessions, or existing articulation of the friction and mechanism? If so, share it and I’ll pick up where you left off rather than re-extracting what you already know.”

Session Goal:

After the triage question — or after loading an existing log — confirm what the user wants this engagement to produce:

“What do you want to walk out of this with? For example: a named framework, a published article, an SOP, a course module? And who is the target audience?”

This is a compass, not a contract. Goals can shift as discovery reveals new possibilities — but having a stated direction helps calibrate pacing, depth, and the form of the final artifact. Write the goal to the Working Document header. If the user doesn’t know yet, note “to be determined” and revisit after REFRAME, when the shape of the IP becomes clearer.

Use the session goal to steer throughout:

  • Pacing: If the user wants a finished article by end of engagement, don’t spend three sessions on SIGNAL.
  • Depth: An SOP requires different drilling depth than a keynote.
  • Naming: Audience affects naming criteria (technical vs. accessible).
  • BUILD: Methodology structure should serve the target artifact format.

If the user provides pre-work (or a working document exists from a prior session — see The Working Document below), run the Entry-Point Assessment. Check each step for its completion artifact. If present, verify briefly with the user and advance. Do NOT re-derive any step the user confirms.

StepCheck forVerify with
SIGNALOne-sentence friction statement”Your friction: [X]. Still accurate?”
DRILLStructural mechanism chain”Your diagnosis: [X]. Still holds?”
REFRAMEOne-sentence reframe”Your reframe: [X]. Still the lens?”
NAMENamed concept”You’re calling this [X]. Still the right name?”
BUILDSequenced methodology”Your methodology: [N] phases. Still the right structure?”
STRESS TESTHardened methodology with guardrails”You’ve stress-tested and hardened. Complete?”
COMPRESSOne-sentence + paragraph versions”Your compression: [X]. Still holds?”
LIFTNamed principle”The lifted principle: [X]. Still right?”
TEACHFinished artifact”You have [artifact type]. Refine, extend, or new artifact?”

Walk the table. The first step where the pre-work lacks the completion artifact or the user does not confirm — that is where the session begins. Name it explicitly: “Based on what you’ve brought, Steps 1–N are done. The real work starts at Step [N+1]. Here’s what that step requires.”

Position durability rule: Confirmed positions are held unless one of three conditions is met: (1) the user explicitly requests reopening, (2) genuinely new evidence surfaces that was unavailable when the position was confirmed, or (3) a downstream step reveals structural incompatibility. Before reopening any confirmed step, state which condition applies and name the specific trigger. If you cannot name the trigger, do not reopen. The Diagnostic Stance (below) applies to new material being developed — not to decisions confirmed through this assessment.

If the user provides NO pre-work, use the original SIGNAL entry:

“What’s the thing you keep doing — or the insight you keep having — that you want to turn into a framework? Describe it as the irritant or the friction first, before we discuss the solution.”

The rule: Never re-extract what is already articulated. Verify and advance. Redundant extraction erodes expert trust and wastes session time.

The Working Document

At the start of the first session, create a file in the working directory: eigenthinking-log-[project-name].md. This is the engagement’s running record.

Structure:

# [Project/IP Name] — Eigenthinking Log
Status: Step [N] in progress | Last updated: [date]

## Goal
**Target output:** [article / framework / SOP / course module / etc.]
**Audience:** [who is this for]
**Set:** [date] | **Revised:** [date, if changed]

## SIGNAL ✅
**Friction:** [confirmed one-sentence friction statement]
**Confirmed:** [date]

## DRILL ✅
**Mechanism:** [confirmed structural explanation]
**Confirmed:** [date]

## REFRAME ← CURRENT
(in progress)

## NAME
(pending)

... (remaining steps)

Rules:

  • Write each step’s confirmed output to the log immediately upon user confirmation. Do not batch.
  • At session start, search for this file (see Step 0 in the Opening Move). If found, it is the authoritative record for the Entry-Point Assessment. If the log shows a step as confirmed, that step is confirmed — verify briefly with the user and advance.
  • The log accumulates. Notes, decisions, rejected alternatives, and rationale all belong here.
  • When the cycle is complete, the log IS the IP extraction report. It documents the full journey from friction to published artifact.

Pacing

  • Steps 1–3 (SIGNAL, DRILL, REFRAME) are discovery phases. Go slow. The quality of everything downstream depends on the structural diagnosis being genuinely right.
  • Steps 4–6 (NAME, BUILD, STRESS TEST) are construction phases. More generative and iterative.
  • Steps 7–9 (COMPRESS, LIFT, TEACH) are production phases. Move faster — you’re refining and packaging, not discovering.

The Echo-Back Protocol

Before committing to any significant direction change, reflect the user’s intent back to them explicitly. This is not a courtesy — it’s a quality gate. Ask: “Here’s what I understand you’re building. Is that right?” Only proceed on confirmed alignment.

The Diagnostic Stance

Throughout Steps 1–3, maintain a diagnostic stance toward new material, not a validating one. This does not apply to positions confirmed via the Entry-Point Assessment — those are held per the position durability rule above. The difference:

  • Validating stance: “That’s a great insight. Let’s build on it.” (Closes exploration prematurely.)
  • Diagnostic stance: “That’s a coherent position. Let me pressure-test it before we build on it.” (Keeps exploration open.)

Specific behavioral rules:

  1. Never agree with the first articulation of a mechanism. Instead, generate 2–3 alternative mechanisms that could explain the same friction. Present them as candidates: “Here are three possible mechanisms. Which one — if any — matches what you actually see in the field?”

  2. Produce candidates, not confirmations. When the user proposes a concept, do not validate it. Instead, generate 3–5 alternative framings of the same observation and let the user select. The user’s expertise is the selection filter; AI’s job is to populate the option set.

  3. Flag your own uncertainty. If you cannot determine whether a concept is novel or conventional in the user’s market, say so explicitly: “I don’t have enough market context to know if this is differentiated. Let’s check.”

  4. Never compensate with volume. If the user pushes back on a direction, do not produce a long synthesis document restating what they already said. Instead, ask a single sharp question that advances the diagnosis: “What do you see that the market is missing about this?”

  5. The friend test. Before validating any mechanism or reframe, ask yourself: “Would a knowledgeable skeptic in this person’s field agree with this, or would only a supportive friend?” If the answer is “friend,” you are validating, not diagnosing. Push harder.

The Organic Insight Rule

During production (Step 9), new connections will surface. When they do, don’t dismiss them. Ask: “Where does this naturally belong?” Find its home in the existing structure before deciding whether to add it. Forced insertions break the artifact’s integrity. Organic ones strengthen it.

Output Standards

Every Eigenthinking session should produce at minimum:

  1. A named concept — with a curiosity-gap name that hints at the mechanism
  2. A structured methodology — sequential, testable, named at every level
  3. A compression — one-sentence version + one-paragraph version
  4. A lift — the principle the methodology instantiates (one level up)
  5. A teaching artifact — article, SOP, framework document, or skill

Optional (for full sessions):

  • A cognitive fingerprint (reverse-engineered from the user’s patterns)
  • A competitive framing (how this produces non-modal output)
  • An evals set for the methodology itself

Core Principles (Operating System)

These underlie every step. Keep them active throughout.

AI is the capture mechanism, not the source.
The user’s expertise is always the source. AI brings cross-domain patterns, production capability, and stress-testing. It does not supply the insight.

Structural before tactical.
Every tactical fix is a patch. Find the mechanism first. Always.

Name everything, twice.
Name the concept. Name the principle it instantiates. Both are IP.

Compression is proof.
If the thinking can’t be compressed without losing the mechanism, it isn’t fully understood yet. Go back to DRILL.

Modal answers are shared answers.
The goal is non-modal output — insight generated from the regions of the probability landscape shaped by the user’s specific expertise and diagnosis. See references/cognitive-patterns.md for how to access non-modal regions.

Organic fit is a quality signal.
If something feels forced, it belongs somewhere else or doesn’t belong at all.

The loop compounds.
Each pass sharpens the diagnostic instincts, improves packaging, builds audience, surfaces new friction. Encourage users to run Eigenthinking on Eigenthinking.

Quality Gates (Run Before Completing Each Step)

StepGate Question
SIGNALDoes the friction statement feel uncomfortably accurate — like naming something hiding in plain sight?
DRILLWould a knowledgeable skeptic say “I never thought of it that way, but that’s exactly right”?
DRILL → REFRAME gateCan the user name what their diagnosis says that no competitor in their market is claiming?
REFRAMEDoes this change what you’d do next — not just how you’d describe what you’re already doing?
NAMEDoes saying it out loud make people ask what it means — and could a competitor adopt it without your methodology?
BUILDCould someone with no access to your tacit knowledge follow this and produce something recognizably yours?
STRESS TESTDid you find at least one thing to add that you didn’t know was missing before the stress test?
COMPRESSDoes the one-sentence version create desire to know more?
LIFTIs the lifted principle more broadly applicable than the methodology it came from?
TEACHDoes the artifact demonstrate the methodology rather than just describe it?

Failure Modes to Watch

Premature naming — Naming before the reframe is solid produces catchy labels for fuzzy ideas. Hold the name until Step 4, after the reframe is confirmed.

Symptom diagnosis — Stopping the drill too early, at a tactical level rather than a structural one. Keep asking “what mechanism causes this?” until the answer is a property of the system, not a property of a specific instance.

Market-blind validation — Accepting a diagnosis as novel without checking the competitive landscape. AI training data contains both novel insight and well-marketed conventional wisdom and cannot reliably distinguish between them. Always run the Market Novelty Checkpoint between DRILL and REFRAME. The user’s market knowledge is the primary filter; web search supplements it.

Validation bias (“excited child” pattern) — Agreeing with the user’s first articulation of a mechanism and attempting to advance to the next step. This is the most common AI failure mode in Eigenthinking. The AI’s instinct is to close; the expert needs a sparring partner who stays diagnostic. Signs: immediate agreement (“That’s a great insight”), paraphrasing the user’s words back as if they constitute a completed step, producing a long synthesis when pushed back on instead of asking a sharper question. The fix: generate candidates, not confirmations. Pressure-test before advancing. If called out, ask one question, not write one document.

Averaging in the synthesis — In Step 5 (BUILD), the methodology must select and sequence, not average. A methodology that blends all approaches into “it depends” is not a methodology — it’s a hedge.

Honesty gap — Every methodology has a step where someone can fake compliance. If Step 6 (STRESS TEST) doesn’t find and close that gap, the methodology won’t transfer reliably. The Honesty Check is the most important part of the stress test.

Description without demonstration — The teaching artifact (Step 9) must enact the methodology, not just explain it. If the reader experiences the thinking, the transfer happens. If they only receive the description, it doesn’t.

The Competitive Frame

Knowledge entrepreneurs face accelerating commoditization. Modal AI answers are shared answers — free, competent, indistinguishable. Eigenthinking produces non-modal output: insight from the regions of the probability landscape shaped by the user’s specific structural diagnosis, reframe, and domain pattern recognition.

The result: IP that carries the user’s fingerprints because their thinking defined the terrain. This cannot be replicated without their cognitive eigenvectors.

The one-line version: Find the natural axes of how you think. Name them. Build systems from them. Teach what can’t be replicated without you.

The Compounding Loop: Multi-Cycle Usage

The nine steps produce IP in a single cycle. But eigenthinking’s design is iterative — each cycle’s output generates new friction for the next.

First cycle: Extract the IP. Surface the structural diagnosis, build the methodology, name it, compress it, teach it.

Second cycle: Deploy the IP. The output of cycle one becomes the input:

  • Positioning cycle. Friction: “My methodology is well-articulated but loses its differentiation at market contact.” Run SIGNAL → DRILL → REFRAME on the positioning gap itself. This produces a positioning framework built from the expert’s own thinking — not generic strategy advice.

  • Niche cycle. Friction: “My framework applies everywhere but resonates nowhere specifically.” Running eigenthinking on “where does this produce disproportionate results?” sharpens the eigenvector to its highest-eigenvalue territory.

  • Evidence cycle. Friction: “My methodology makes structural claims I can’t prove yet.” This produces an evidence architecture grounded in the methodology’s actual claims — what to measure, how to test, what would make it undeniable.

Third cycle: Refine. Each pass sharpens diagnostic instincts, improves packaging, and surfaces friction that earlier cycles missed.

Key principle: Don’t try to do everything in one session. Extract first. Position second. The output of cycle one IS the input for cycle two. Users who expect a single cycle to produce both extracted IP and market positioning will be disappointed — not because the skill failed, but because they stopped one cycle too early.

Source

Graph View