# The AI Method Lexicon

100 method names a model already knows, sorted by the job you need done, plus 12 commands that stack them into workflows.
By Drew Brucker. The live version, with real before-and-after runs: https://drewcbrucker.com/goodies/ai-conceptual-zip-files/

## How to use a name

Drop the name into your prompt, then say exactly what you want back. The name primes the method, and your instructions are what make it repeatable.

Each method carries a type:

- **P · Procedural:** procedural and usually easy to specify.
- **L · Lens:** a broad intellectual lens that benefits from explicit output instructions.
- **D · Data-dependent:** data-dependent and may be unreliable without sources, measurements, or computation.

## Stress-test

Attack the idea before someone else does.

- **Red teaming** (P): Stress-test a proposal from an intelligent adversary’s perspective.
- **Pre-mortem** (P): Assume failure; reverse-engineer its most likely causes.
- **Steelmanning** (L): Build the strongest defensible opposing argument.
- **Falsification** (P): Seek observations that would disprove a claim.
- **Inversion** (L): Work backward from failure or an unwanted outcome.
- **Dialectical inquiry** (P): Test a thesis against an explicit antithesis.
- **Double crux** (P): Find the precise disputed belief driving disagreement.
- **FMEA** (P): Catalog failure modes, effects, severity, and mitigations.
- **Fault-tree analysis** (P): Trace how combinations of failures cause a bad event.
- **Disconfirmation search** (P): Seek high-quality contrary evidence before committing.
- **Socratic questioning** (P): Probe definitions, evidence, implications, and alternatives.

## Diagnose

Find the real cause behind the symptom.

- **First-principles reasoning** (L): Rebuild a solution from underlying constraints and facts.
- **Theory of Constraints** (P): Find and improve the limiting system bottleneck.
- **Five Whys** (P): Iteratively probe the mechanism behind a recurring issue.
- **Causal-loop diagramming** (P): Map reinforcing and balancing feedback loops.
- **Abductive reasoning** (L): Find the most plausible explanation of limited evidence.
- **Pareto analysis** (P): Find the minority of causes driving most outcomes.
- **Ladder of inference** (P): Trace how observations become interpretations and beliefs.
- **Causal DAG** (D): Represent candidate causes, confounders, and effects.
- **After-action review** (P): Compare intent, reality, causes, and next changes.
- **Critical-incident technique** (P): Study extreme memorable episodes to identify causes.

## Decide

Make the call when the future won't hold still.

- **Assumption mapping** (P): Rank beliefs by importance and uncertainty.
- **Sensitivity analysis** (P): Show which inputs could reverse a conclusion.
- **Second-order thinking** (L): Trace downstream effects, reactions, and feedback.
- **Reference-class forecasting** (D): Estimate outcomes from genuinely comparable cases.
- **Expected-value analysis** (P): Compare probability-weighted gains and losses.
- **Scenario planning** (P): Prepare for several plausible, meaningfully different futures.
- **Counterfactual analysis** (L): Change a premise and examine resulting consequences.
- **Cynefin framework** (P): Choose action style based on problem complexity.
- **Bayesian updating** (D): Revise beliefs explicitly when new evidence arrives.
- **Decision-tree analysis** (P): Map alternatives, contingencies, probabilities, and outcomes.
- **Value of information** (P): Decide which uncertainty is worth resolving first.
- **Real-options thinking** (L): Prefer investments that preserve valuable future choices.
- **Backcasting** (P): Work backward from a specific future end state.
- **Monte Carlo simulation** (D): Model outcome distributions from uncertain inputs.
- **Fermi estimation** (P): Decompose a rough estimate into checkable pieces.
- **OODA loop** (P): Observe, orient, decide, act, then repeat.

## Invent

Get past the first answer everyone else lands on.

- **Analogical transfer** (P): Import mechanisms from a dissimilar domain.
- **Conceptual blending** (P): Fuse two conceptual systems to produce fresh ideas.
- **Lateral thinking** (L): Escape default frames through provocative alternatives.
- **Morphological analysis** (P): Combine options across independent problem dimensions.
- **TRIZ** (P): Resolve design contradictions using inventive principles.
- **SCAMPER** (P): Generate ideas via seven structured transformation verbs.
- **Synectics** (P): Use metaphor and distant analogy to spark ideas.
- **Assumption reversal** (P): Invert category conventions to expose new possibilities.
- **Biomimicry** (L): Borrow mechanisms from natural systems and organisms.
- **Design fiction** (P): Explore futures through plausible fictional artifacts.
- **Speculative design** (L): Create provocations about possible alternative futures.
- **Constraint-led ideation** (P): Use deliberate limitations to force different solutions.
- **Divergent–convergent thinking** (P): Generate broadly, then select deliberately.
- **Design sprint** (P): Time-box framing, sketching, selection, and testing.
- **Double Diamond** (P): Separate discovering/defining problems from developing/delivering solutions.

## Position

Find where you win and why it holds.

- **Jobs to Be Done** (P): Identify the progress customers hire a product to make.
- **Opportunity-solution tree** (P): Connect desired outcome to opportunities, ideas, experiments.
- **Semiotic analysis** (P): Decode cultural signs, visual cues, and their meanings.
- **Category design** (L): Define and make legible a distinctive market category.
- **Kano model** (P): Sort features by basic, performance, and delight effects.
- **Blue Ocean four-actions grid** (P): Eliminate, reduce, raise, and create value factors.
- **Wardley mapping** (D): Map user needs, value chains, and component evolution.
- **7 Powers** (L): Analyze sources of enduring competitive advantage.
- **Porter’s Five Forces** (P): Assess industry pressures on profit potential.
- **Value-chain analysis** (P): Locate where an organization creates and captures value.
- **Game-theoretic analysis** (L): Model incentives and likely strategic responses.
- **Ecosystem mapping** (P): Map complementors, intermediaries, dependencies, and flows.
- **Stakeholder mapping** (P): Surface actors by influence, incentives, and interest.
- **Strategic-group mapping** (P): Compare competitor clusters along two meaningful dimensions.
- **Network-effects analysis** (P): Assess how added participants change product value.
- **Flywheel mapping** (P): Identify reinforcing activities that improve with repetition.
- **Growth-loop modeling** (P): Trace how an output feeds the next acquisition cycle.
- **Working Backwards / PR-FAQ** (P): Clarify a proposed offering through a future announcement.

## Research

Get evidence you'd bet on, then test it.

- **Triangulation** (P): Check claims using independent methods or sources.
- **Lead-user method** (D): Learn from advanced users facing needs early.
- **Systematic review** (D): Search and synthesize evidence with explicit inclusion rules.
- **Evidence grading** (P): Weight evidence by quality, relevance, and uncertainty.
- **Design of experiments** (D): Vary factors systematically to estimate effects.
- **Difference-in-differences** (D): Compare outcome changes across treated/control groups.
- **Interrupted time-series analysis** (D): Estimate impact around an intervention from trend data.
- **Cohort analysis** (D): Compare groups by entry period or shared experience.
- **Conjoint analysis** (D): Infer preference tradeoffs from structured choice data.
- **MaxDiff analysis** (D): Estimate relative priorities from best–worst choices.
- **Thematic analysis** (P): Code qualitative material to identify recurring patterns.
- **Grounded theory** (D): Build explanatory concepts iteratively from qualitative evidence.
- **Discourse analysis** (P): Examine how language constructs meaning and power.
- **Delphi method** (D): Iteratively aggregate independent expert judgments.
- **Cognitive task analysis** (P): Unpack expert decisions, cues, and hidden knowledge.
- **Contextual inquiry** (P): Observe users doing work in their real environment.
- **Cognitive walkthrough** (P): Step through user tasks to locate confusing interfaces.
- **Metamorphic testing** (P): Test whether changed inputs preserve expected relationships.
- **Blind evaluation** (P): Judge alternatives without brand/model/source labels.
- **Heuristic evaluation** (P): Check designs against established usability principles.
- **Laddering interviews** (P): Connect attributes to benefits, motives, and values.

## Communicate

Say it so the right person acts on it.

- **MECE decomposition** (P): Partition a problem into nonoverlapping, covering categories.
- **Pyramid Principle** (P): Lead with the conclusion, then supporting logic.
- **Argument mapping** (P): Diagram claims, support, objections, and rebuttals.
- **Feynman technique** (P): Explain simply and locate gaps in understanding.
- **BLUF** (P): State the bottom line before supporting detail.
- **Inverted-pyramid writing** (P): Order information from most to least important.
- **Story circle** (P): Structure character change through eight narrative beats.
- **AIDA** (P): Sequence attention, interest, desire, and action.
- **Minimum-description-length principle** (L): Prefer the simplest explanation that compresses the data.

## The 12 commands

Syntax: `/command [target]; [objective]; [constraints]`

Names are user-defined, not model-native commands. They require the rules below to be available in the current model’s context, project, or skill.

### /redteam: A critical adversarial review

- **Methods:** Steelmanning, falsification, pre-mortem.
- **Protocol:** State the proposal and success criteria. Build the strongest plausible objection. Identify failure modes and contradictory evidence. Rank weaknesses by impact × likelihood, then propose fixes.
- **Output:** Verdict; 3-5 critical risks; evidence and uncertainty; prioritized repairs.

### /xray: Root-cause diagnosis

- **Methods:** Five Whys, causal mapping, Theory of Constraints.
- **Protocol:** Describe observed symptoms separately from interpretations. Form rival causal hypotheses. Map dependencies and find likely constraints. Recommend the smallest discriminating tests, then the most effective intervention.
- **Output:** Root-cause hypotheses; evidence gaps; bottleneck; test; intervention.

### /asymmetry: High-upside opportunity search

- **Methods:** Analogical transfer, expected value, real options.
- **Protocol:** Define resources, constraints and desired upside. Generate opportunities across known and adjacent markets. Evaluate upside, probability, downside, effort, time-to-signal, defensibility, and option value. Recommend a staged low-cost experiment.
- **Output:** Ranked bets; core assumptions; fastest proof; stop criteria.

### /white-space: Distinctive market territory

- **Methods:** JTBD, semiotics, category design, Blue Ocean.
- **Protocol:** Map incumbent language, visual codes, benefit claims and customer jobs. Find crowded and underserved territories. Develop positions based on a true capability and meaningful customer difference. Pressure-test against competitors.
- **Output:** Market map; 3 distinct territories; preferred position; proof needed.

### /crosswire: Cross-domain creative invention

- **Methods:** Analogical transfer, conceptual blending, TRIZ.
- **Protocol:** Select 4-6 distant disciplines. Import specific mechanisms rather than aesthetics alone. Blend with the target brief. Explain why each idea works, what makes it original, and what it would require to execute.
- **Output:** 6 genuinely different concepts; transferable mechanisms; top 2 picks.

### /territories: Campaign or brand concept exploration

- **Methods:** Morphological analysis, divergent-convergent ideation, semiotics.
- **Protocol:** Define core message and audience. Vary narrative mechanism, visual grammar, channel expression, emotional register, and cultural reference. Create distinct territories, not minor style variants. Rank against the brief.
- **Output:** 5-7 territories with hook, premise, visual rules, sample execution and ranking.

### /taste-lock: Forensic aesthetic direction

- **Methods:** Semiotic analysis, comparative evaluation, exclusion rules.
- **Protocol:** Extract what is consistently loved and disliked from supplied references and feedback. Separate invariant constraints from movable choices. Turn subjective taste into testable design rules, anti-patterns, and a critique rubric.
- **Output:** Taste DNA; do/dont rules; 5-axis rubric; example evaluations.

### /signal: High-signal current research

- **Methods:** Triangulation, evidence grading, disconfirmation.
- **Protocol:** Define decision-relevant questions and recency. Use current primary evidence where tools allow. Seek practitioners and independent corroboration. Distinguish confirmed facts, inferences, predictions and marketing claims.
- **Output:** What changed; why it matters; confidence; primary sources; action.

### /package: Content discovery and packaging

- **Methods:** JTBD, positioning, curiosity-gap analysis, blind evaluation.
- **Protocol:** Use the provided episode or transcript as the source of truth. Identify the strongest viewer promise and search intent. Propose contrasting title lanes, thumbnail hooks, description and chapters only when timestamps are given. Never invent quotes or timestamps.
- **Output:** Target queries; differentiated titles; strongest hybrid; thumbnail angle; SEO copy.

### /voice: Personal voice rewrite

- **Methods:** Feynman technique, vivid concrete language, BLUF.
- **Protocol:** Preserve original meaning and confidence level. Write in direct, accessible language with a natural casual cadence, restrained slang and well-chosen vividness. Remove jargon, boilerplate, over-explaining, forced jokes and em dashes.
- **Output:** Polished version; optionally 2 materially different alternatives; only key tradeoffs.

### /decision: Explicit choice architecture

- **Methods:** Expected value, sensitivity analysis, pre-mortem, real options.
- **Protocol:** Identify decision, objectives and hard constraints. Compare realistic options. Separate evidence from estimates. Stress test the conclusion on the two most sensitive assumptions. Recommend one move with reversibility and decision triggers.
- **Output:** Ranked decision; decisive criteria; risks; what would change the call.

### /ship: Final quality gate

- **Methods:** Rubric evaluation, adversarial checking, checklists.
- **Protocol:** Audit the deliverable against the brief, truthfulness, brand, format, relevance, and operational readiness. Flag anything missing or unverified. Fix deficiencies when possible, and distinguish completed checks from checks that require human or tool verification.
- **Output:** Pass/fail by criterion; fixes made; unresolved issues; final release recommendation.

## House rules for every command

- An invoked mode determines *method*, never permission to ignore the user’s specific task, sources, scope, or other applicable instructions.
- Do not invent facts, sources, quotations, timestamps, metrics, research, testing, or live verification.
- Identify uncertainty and assumptions when material. Run external checks when needed and tools are available; otherwise describe what was not verified.
- Keep procedures proportionate to the task; do not perform unnecessary elaborate analyses.
- If two modes are invoked, apply the first to generate or diagnose and the second to evaluate or finish, unless the task specifies the reverse.
- User instructions override default output templates. Do not reveal hidden/private reasoning; give concise decision rationales and evidence.

## Examples

- `/white-space: Position my consulting offer for high-end in-house creative teams. Prioritize distinctiveness and provable outcomes.`
- `/crosswire + /territories: Develop concepts for a premium AI design platform. Avoid generic futuristic imagery.`
- `/package: Work from the supplied Fast Hours transcript. Give me two title lanes, search queries, and a hybrid pick. No invented timestamps.`
- `/redteam + /decision: Is a podcast sponsorship product worth prioritizing over a new consulting offer?`
- `/taste-lock: Derive the invariant rules from these 12 references and my feedback.`

## A note on reliability

A single word can prime a known methodology, but it does not reliably specify a repeatable procedure across models. Stable definitions, suitable context, examples, tools and evaluation make the shorthand stronger.
