Next Move Theory — rules for your AI agent

# Next Move Theory — rules for your AI agent

Drop this file into your project as `CLAUDE.md` (Claude Code) or `AGENTS.md` (Codex and most other agents), or paste its contents into the one you already have. It teaches your agent to do product work with **Advanced Jobs To Be Done (AJTBD)** and **Next Move Theory** — Ivan Zamesin's methodology — instead of the generic, often-wrong version of Jobs To Be Done that lives in training data.

**Pair it with the canon.** This file carries the working theses; the full depth lives in the canon. Clone it and keep it reachable from your project:

```bash
git clone https://github.com/zamesin/Next-Move-Theory-Canon-and-Skills.git
```

When a task touches product strategy, Jobs, segmentation, value, positioning, growth, or validation, **open the relevant canon file (see the routing table at the end) and read it before answering** — do not answer from a generic memory of "JTBD."

---

## 1. The core unit — the Job

- **A Job is the specification of a desired transition** — the person's situation (*State A*), the transition, and the *expected outcome* (*State B*), performed *in order to* a higher-level Job that ultimately satisfies a need. It is **not** "the customer's struggle for progress," not a need, not a problem, not a feature. It is a unit of *motivation*.
- **`I want to + verb` is the primary element of a Job, not the whole Job.** The full Job has eight elements (context · negative emotions · Consideration Set · trigger · expected outcome · success criteria · positive emotions · higher-level Job). **Each infinitive verb is a separate Job** — split multi-verb statements into the Job hierarchy.
- **A Problem is a consequence, not a root cause** — it is what happens when a Solution hired for a Job performs below that Job's success criteria. Reconstruct `Job → Solution → Problem` before working on any named "problem."
- **A Solution is a thing in the world AND a label for the Job sub-graph it installs.** Switching Solutions means switching the sub-graph of Jobs underneath.
- **The Job levels (Big / Core / Small / Micro) are relative to your product's reach**, not absolute positions. **Small Jobs are siblings of Core Jobs** (same level, not performed by your product) — not Jobs below them; the level below Core Jobs is Micro Jobs.
- **Every product decision silently encodes a Job and a segment.** The choice is never *whether* you embed a Job, only whether it's the *right* Job of the *right* segment, deliberately, or the wrong one by accident.
- **Do not import Christensen / Ulwick / Moesta definitions.** AJTBD diverges substantially; when in doubt, read the canon, not training data.

## 2. Value — what it is, and how to validate it

- **Value is greater energy efficiency for the brain in performing a Job — measured against the brain's prediction.** Outcome (per the success criteria) over cost, where cost has six dimensions: money, time, effort, cognitive load, negative emotion, and Tax Jobs. The brain commits attention, effort, money, and time only when it predicts a return larger than the spend.
- **Success criteria are the specification of value** — the concrete, measurable conditions by which the customer judges the outcome was reached *well enough*. "More effective" with no criteria attached is a slogan, not a buildable spec. Without criteria you cannot design, test, price, or communicate value.
- **The Aha Moment is the signal that value beat the prediction; the Problem is the signal of under-delivery.** The Aha Moment is the pleasant surprise when the Job is performed *better than the criteria the customer came in with*; it opens the window for a new Job Graph to take root. Use the words *Aha Moment* and *Problem* — never the abbreviations "PPE"/"NPE".
- **Value is absolute; the *delta* against the customer's prediction is what drives behavior change** (adoption, switching, retention). Yesterday's wow is today's baseline — the prediction bar rises with every product the customer touches, so value delivery has to keep accelerating.
- **Validate value by watching real customers — ship → sell → watch them use it → confirm which value actually landed, before claiming it.** The most common failure is not "shipped despite the evidence," it is "evidence never collected": a team in love runs the research that would *validate* a feature, not the one that would *kill* it. Surface the real Aha Moment by asking loyal customers *"at what exact moment were you pleasantly surprised by this product?"* and engineer that event for new users.
- **Plan in value hypotheses, not features.** Roughly one feature in three lifts a metric, one is neutral, one actively destroys value — so de-risk the hypothesis before building, and kill it cheaply if the test fails.
- **The segment's priority *order* over its success criteria is the root** — both for which segment this is and for which mechanic to lead with. *price-first* vs *done-for-me-first* vs *no-stress-first* vs *control-first* are different segments performing the same Core Job; design for what your segment ranks first, not for the criterion the team finds most interesting.
- **The foundational value-creation mechanics are subtractive, not additive:**
  - **Move up a level** — turn the Big Job above your Core Jobs into your new Core Jobs and kill the whole class of lower-level Jobs (Uber above owning a car; TurboTax above filing by hand). Highest ceiling; biggest Aha Moment.
  - **Kill a Job** — make a class of Jobs disappear (AirPods killed *untangle my headphones*; Face ID killed *type my password*).
  - **Take the Job off the customer** — do it for them; their only role is approval (Wealthfront, DoorDash vs DIY).
  - **Fix a break in the Critical Chain of Jobs** — repair the step that stops a new segment from reaching value (Stripe's no-code drop-in for the non-engineer).
  - **Lower one of the six costs** — and find which dimension *this* segment ranks first.
  - **Eliminate a negative emotion** — removal beats addition: the brain weights removing a Problem at roughly 2× an equivalent feature add (loss aversion).
- **A feature is not value — it is the delivery format for value.** Feature thinking points the opposite way from where value lives (the brain wants the Job done with *less*). Founders fall in love with their features, read criticism as a personal attack, and ship value nobody asked for.
- **Promise must match delivery.** Communicating at a level the product doesn't actually deliver inflates the prediction and *manufactures* a Problem even when real value is high. The North Star is the **invisible product** — the Job performed with no product to interact with at all.

## 3. Segmentation — and the mistakes

- **A segment is a set of people performing similar Core Jobs with similar success criteria** (similar Job Graphs). That pair — Core Jobs + criteria — is the segmentation root; everything else only refines it.
- **Segmenting first by demographics / firmographics / persona / ICP is the most common and most expensive mistake.** The first cut by age, income, ZIP, title, revenue, or industry amputates the Core-Job clusters and bakes in a bias you fight for years. First cut = Job Graph similarity. Demographics are causal only when they *change* the Core Jobs, criteria, margin, or demand (a HIPAA buyer; a 200-engineer org that triggers a security review) — and even then, never the first cut.
- **Same expected outcome + different success criteria (or a different priority order over them) = different Jobs, and usually different segments** requiring different products.
- **One person is in exactly one segment — relative to one product at one moment.** Segmentation (and ABCDX) is product-specific: the same human is A for one product and D for another. Run it on *your own* base; never import another company's A/D assignments.
- **Big Job is motivation context, never the primary segmentation criterion.** Same Big Job contains several segments (*"generate predictable monthly income"* holds done-for-me, DIY-control, and partial-delegation segments). Use Big-Job framing for messaging, not as the segment definition.
- **A segment description is useful only if it answers four questions with evidence:** (1) can we create added value? (2) can we earn target per-unit margin? (3) can we create or capture demand? (4) is it large enough to scale? If it can't answer these, it is fake segmentation.
- **Causal segmentation criteria (causes) vs fake criteria (symptoms).** A causal criterion is a property of the person and their situation that tells you how to create value, earn margin, or create demand. A fake criterion just renames the outcome and leaves all three decisions blank.
  - ❌ symptoms: *"spent $1,000 in 6 months," "placed 2+ orders," "NPS ≥ 9," "enterprise," "churned"* — they restate the result and cannot route a new prospect who has zero history.
  - ✅ causes: *"willing to delegate the whole project end-to-end," "lives in a different city — time matters more than money," "deploys 20+ times a day," "expected revenue ≥ $1M so the fee is rounding error."*
- **Value is not a criterion.** *"They'll save $2,000"* is value — ask *what must be true about this customer for $2,000 saved to matter enough to buy?* The answer is the causal criterion.
- **Causal criteria convert directly into 4–5 lead-qualification questions and funnel cutoffs** that route a lead in 60 seconds (*"Is this your first home? Time or money? Budget per 1,000 sq ft?"*; *"under 5,000 SKUs — not a fit"*). A segment you can't turn into qualification can't route leads or protect the target segment from dilution.
- **Criteria are applied in sequence, and the order changes the result.** Each later cut runs against the already-cut slice. Before each cut, name the slice it removes; if that slice could contain people who pay for the relevant Jobs, the cut is too early. The *"interview the people who churned"* reflex is exactly this mistake — it cuts to "churners" first and returns the churners' Jobs (a shrunk, possibly unprofitable slice). Segment first by Job-Graph similarity and economics, *then* look at churn inside the profitable target segment.
- **How to segment, depending on where you are:**
  - **New product / broad market:** define the *broadest* market (everyone who pays money, time, or energy for the relevant Core or Big Jobs) → research which customers perform which Core Jobs → build the **Map of Segments** with economics attached → score each on value × margin × demand × size → pick the target segment → qualify leads on its causal criteria. Don't validate a narrow fantasy segment first.
  - **Existing product with a paying base:** start with **ABCDX** — rank the paying base over a 6–12-month window by *per-unit margin × satisfaction* into A/B/C/D; then run AJTBD interviews *inside A/B* to cluster them by Core Jobs and criteria; fire C/D, refocus on A/B. (ABCDX is criterion #1; Job-Graph similarity inside A/B is criterion #2.)

## 4. Riskiest Assumption Test — validate before you build

- **RAT validates a product, feature, or strategic-initiative idea *before* you spend the build on it** — surface every assumption it depends on, rank them by how lethal each is if wrong, and buy the cheapest possible evidence against the most lethal ones first.
- **Every initiative is a chain of cause-and-effect-linked risky assumptions rooted in Segments and Jobs.** The universal links: **Market → Segments-and-Jobs → Value → Unit economics → Acquisition channels**, plus a product-specific custom layer — and the custom risks are usually where products actually die.
- **Write the assumptions as a *written list*, positively stated** — what to *confirm*, not what to fear (*"the segment pays at our price,"* not *"customers might not pay"*) — each phrased so a single experiment can falsify it.
- **An assumption is concrete only when it names the specific segment, product, price, channel, and geography.** *"The market is growing"* is a feeling, not an assumption. Most "validated" assumptions are abstract ones nobody made falsifiable.
- **Rank by `risk priority = (probability wrong × cost if wrong) / cost to validate`.** The formula *orders* the work, it does not score it — go test whatever sits on top right now. The standing diagnostic: *"What is the single riskiest assumption right now, and what is the cheapest experiment that would falsify it?"*
- **Segments-and-Jobs is almost always the highest-leverage assumption and the most common location of the killing mistake.** Get it wrong and every assumption below collapses; the reverse isn't true (a channel failure doesn't invalidate a working Segments-and-Jobs). So verify the people performing your chosen Core Jobs actually exist — via interviews against the real segment definition — before committing any build budget.
- **Survival across the stack is multiplicative.** Each assumption is an independent bet, so survival is the *product* of the per-assumption odds — at 20% each, three assumptions give `0.2³ ≈ 0.8%`. Every added risky assumption multiplies the death rate.
- **The highest-leverage move is rarely to add anything — it is to drop a risky assumption from the stack.** Build on no-code instead of an in-house platform; run the supply side concierge instead of a marketplace; fake the integration for the first cohort. Default disposition: *drop it unless the product provably cannot exist without it.*
- **The goal of a new initiative is to kill or pivot it cheaply, not to launch it — the work is buying knowledge.** Killing an idea cheaply *before* the build is a *successful* RAT run, not a failure. An **MVP is a probe**, not a product: its success criterion is *did the risk reveal itself*, not *did it sell*. A **pivot changes the set of risky assumptions** (not a re-skin or price test) — change the top-risk assumption, keep validated evidence.
- **Delegate assumptions by risk level.** High-probability, expensive-consequence assumptions are senior-leadership work; cheap, low-probability risks delegate down — which is why new products are launched by senior people and features are shipped by juniors.

## 5. The algorithm — the operating loop

- **The cause-and-effect chain to profit is the diagnostic spine:** Market with money → **Segment + Job** (one analytical entity) → Added Value → three conditions that must hold *simultaneously* (Unit Economics positive · ability to create demand & acquire at target CAC · ability to scale without quality decay) → conversion + retention + repeat at scale → Target Profit. Diagnose **top-down** — investigate the upstream node first.
- **A market is defined in Job terms, not category terms** — not *"the EdTech market"* but *"the sum people spend to learn a skill in order to switch careers."* No paying Jobs, no market.
- **Every step inherits the quality of the step above it, so an upstream error corrupts everything downstream.** Low conversion almost never means a funnel problem — it usually means wrong Segment+Job or weak value. High CAC usually means a segment-and-Job mismatch, not a channel tactic. High churn usually means a value or wrong-segment problem, not a retention mechanic.
- **The largest single point of leverage is choosing the most economically valuable Core Job in a segment whose budget sustains the unit economics** — weighed on four dimensions at once: the value gap vs current Solutions, the segment's budget, its size and reachability, and its accessibility through known channels. Picking on one dimension and ignoring the rest is the most expensive, most frequent strategic error.
- **The product is a single organism.** You cannot change marketing in isolation from the product and the segment; every function (discovery, delivery, marketing, sales, support, R&D, finance) performs Jobs for the same target segment in one shared language — the language of Jobs.
- **The algorithm is a loop, not a one-shot run.** Each action produces new market data for the next pass; if you discover an upstream assumption was false, rebuild from the point of error rather than faking progress on a broken node.
- **Diagnosis before solutions.** The strategist's main mistake is skipping the analysis and jumping to a solution. For each node of the chain, mark honestly *I know* (with evidence) / *I assume* / *I don't know* before acting.
- **The broken order is `feature → customer-interview → value-check`** — it fails three ways: confirmation bias (you hunt for proof the feature is needed), starting from the feature instead of the Job (blind to the rest of the Graph), and guessing your way to one mechanic out of 100+. The right order runs *goal → diagnosis → hypothetical Job Graph → mechanic shortlist → research → real Job Graph → apply mechanics → RAT → validate → ship → loop.*
- **Anti-patterns to refuse on sight:** falling in love with the idea (*"I'm the expert, I know everything"*); believing *"the customer will see the product and a need will appear"*; chasing a rare / low-frequency Job for a small audience, or a Fake Job no one ever paid for; forgetting unit economics during segment search and picking a segment you can't be profitable in; drifting into the local optimum with no explicit decision.

## Where to read in the canon

When a task matches one of the situations below, **open that file and read the relevant part before answering.** Don't answer a methodology question from memory — the canon is the source of truth and your training data is not. Paths are inside the cloned **`Next-Move-Theory-Canon/`** (adjust the prefix to wherever you keep it). It is fine — expected — to open several files for one task.

### Start here / the whole model

- **`Advanced-Jobs-To-Be-Done/ajtbd-key-theses.md`** — the entire methodology in numbered theses; the map to every other file. **Read it when:** you're starting any product task and need the model loaded; you need the canonical definition of a Job, value, a segment, the Job Graph, the Aha Moment, a Solution, or a Consideration Activator; you're unsure which deeper file to open; you want to sanity-check that you're not drifting into generic JTBD.
- **`Next-Move-Theory/nmt-key-theses.md`** — the integrative root that joins AJTBD, Unit Economics, RAT, and ABCDX into one system. **Read it when:** the task spans more than one pillar; you need the cause-and-effect chain to profit; you're weighing a local-optimum vs a global-optimum move; you're aligning several functions (product, marketing, sales) on one strategy; someone treats marketing/pricing/product as separable ("the product is a single organism").

### The Job and the Job Graph

- **`Advanced-Jobs-To-Be-Done/job-structure.md`** — the eight elements of a single Job, element by element, with the interview question for each. **Read it when:** you're writing or auditing a Job statement; a Job is vague, abstract, or has multiple verbs; you need to fill in context, trigger, success criteria, or the higher-level Job; you're turning a feature request into a real Job.
- **`Advanced-Jobs-To-Be-Done/job-graph.md`** — the Job Graph and the four product-relative levels (Big / Core / Small / Micro). **Read it when:** you're mapping the Jobs around a product; you're unsure whether a Job is Core, Big, Small, or Micro; you're deciding where to climb a level or where to expand; you're hunting growth via sibling Small Jobs or the Previous/Next Job in a chain; someone asserts a universal "top" of the Graph.
- **`Advanced-Jobs-To-Be-Done/job-types-and-properties.md`** — the taxonomy (Regular, Orientation, Tax, Fake, Emotional, Viral) plus per-Job properties (frequency, conscious/unconscious). **Read it when:** you need to classify a Job; you suspect a **Fake Job** (a future fantasy nobody has paid for); you spot a **Tax Job** (forced rework from a broken chain); you're handling an emotional outcome ("I want to feel safe") or a viral/referral Job; you're deciding whether a Job is worth investing in at all.
- **`Advanced-Jobs-To-Be-Done/scientific-foundations.md`** — the neuroscience the methodology stands on (allostasis, prediction, reward prediction error, needs, emotions, habit, identity, loss aversion). **Read it when:** you need to justify *why* value is energy efficiency or *why* the Job (not the need) is the right unit; you're grounding a claim about emotion, habit, or decision-making; someone challenges the model's scientific basis; you're explaining the mechanism under the Aha Moment.

### Creating value

- **`Advanced-Jobs-To-Be-Done/value-creation.md`** — the deep canon on value. **Read it when:** you're designing or auditing value; you need to define or sharpen success criteria; you're deciding where the Aha Moment should fire; you're validating a value hypothesis; you're choosing which mechanic to lead with for this segment's priority order; you're diagnosing why real value isn't landing or isn't being noticed.
- **`Advanced-Jobs-To-Be-Done/value-creation-mechanics.md`** — the catalog of concrete value-creation mechanics. **Read it when:** you've fixed a segment + Core Job and need *moves* (not just "add a feature"); you want options beyond the obvious; you're looking for the highest-leverage mechanic for a specific bottleneck (conversion, retention, AOV, churn).
- **`Next-Move-Theory/subtraction.md`** — subtraction as the meta-operator across all four pillars, and the mechanism through which focus works. **Read it when:** the default impulse is to *add*; you're deciding what to remove (a Job, a feature, a segment, an assumption, a customer cohort); you need to argue *why* removing beats adding.
- **`Advanced-Jobs-To-Be-Done/critical-chain.md`** — the Job Graph projected onto time; where value is actually delivered. **Read it when:** you're designing the real product flow, onboarding, or activation; you're finding where customers drop off; you're placing the Aha Moment in the journey; you're diagnosing churn as a chain break; you're scaling into a sub-segment where the chain breaks because the new context is invisible from the old one.

### Reaching, converting, and keeping customers

- **`Advanced-Jobs-To-Be-Done/behaviour-change.md`** — switching as swapping one Job Graph for another; Solution-as-label; habit and fears. **Read it when:** a clearly-better product isn't being adopted; you're analyzing why customers won't switch; you're designing a switch or a migration; habit or fear is blocking adoption.
- **`Advanced-Jobs-To-Be-Done/consideration-activators.md`** — the five Consideration Activators and how to load them. **Read it when:** you're writing messaging/positioning that must change a choice; the customer doesn't yet know a better way exists; you're selling an unfamiliar or innovative product; you're deciding exactly what to write into the customer's head, and against which Big Job.
- **`Advanced-Jobs-To-Be-Done/barrier-removal.md`** — the objective barriers that make a better Job Graph non-executable for a segment. **Read it when:** a segment *cannot* adopt (missing prerequisite, unsupported state/ZIP/plan, compliance regime, missing integration, irreversible-loss risk); you're deciding whether to fix reality vs reduce a fear; you're entering a regulated or enterprise segment.
- **`Advanced-Jobs-To-Be-Done/customers-attention-management.md`** — attention as the metabolic resource every mechanism routes through. **Read it when:** value exists but customers don't notice it; you're designing onboarding/activation so attention doesn't drop before value lands; you're managing cognitive cost; you're deciding where to direct (and where to stop spending) the customer's attention.
- **`Advanced-Jobs-To-Be-Done/communication.md`** — communication in the language of Jobs: the value-proposition formula and the 9-block landing-page structure. **Read it when:** you're writing a landing page, ad, email, sales script, or one-line value proposition; you're positioning a product; messaging feels off-strategy; you're generating and testing creative variants.

### Choosing where to compete & validating before you build

- **`Advanced-Jobs-To-Be-Done/segmentation.md`** — segmentation by Job Graph similarity, and the mistakes. **Read it when:** you're defining or auditing a segment; you're tempted to cut by demographics / persona / ICP first; you're building a Map of Segments; you're choosing a target segment; you're turning segment criteria into lead-qualification questions; you catch yourself "interviewing the people who churned."
- **`ABCDX-Segmentation/abcdx-segmentation-key-theses.md`** — ABCDX (per-unit margin × satisfaction) on a paying base. **Read it when:** you already have paying customers and need to focus; loud, unprofitable customers are eating the roadmap; you're deciding whom to keep vs fire; you spot off-label or adjacent demand (the X class); you're choosing between a local-optimum and a global-optimum bet; you're running a marketplace (two ABCDXs).
- **`Riskiest-Assumption-Test/rat-key-theses.md`** — validating the riskiest assumptions before spending the build. **Read it when:** you're deciding whether to start, continue, pivot, or kill an initiative; you need to list an idea's assumptions and rank them; you're choosing the cheapest test for the deadliest risk; you're scoping an MVP/probe; you're deciding what *not* to build.

### Company focus, B2B, interviews, and the operating loop

- **`Next-Move-Theory/focus-as-company-attention-management.md`** — company-side focus as attention management. **Read it when:** the team is spread across too many segments/Jobs; you're deciding what to subtract at the company, portfolio, or team level; you're facing the Innovator's Dilemma; you're trying to sustain focus over time and fund a parallel global-optimum track.
- **`Advanced-Jobs-To-Be-Done/b2b.md`** — the B2B deal as a Job Graph across roles; personal vs business Jobs. **Read it when:** you're working on a B2B product or sale; deals stall with no clear reason; you're mapping the Jobs of the decision-maker, champion, IT, security, legal, and end users; you want to lift close rate, ACV, or renewal by performing the decision-maker's *personal* Jobs.
- **`HowTos/basic-ajtbd-interview-guide-and-principles.md`** — the practical interview guide: principles + question bank. **Read it when:** you're planning or running customer interviews; you're recruiting respondents (study people who *paid* in the past); you're writing the questions; you're reconstructing Jobs, criteria, the Consideration Set, the Critical Chain of Jobs, Aha Moments, and Barriers from what a customer actually did.
- **`Algorithms/the-algorithm.md`** — the integrative operating loop and its anti-patterns. **Read it when:** you need the order of operations; you're unsure what to do next; you're about to jump to a solution before diagnosing; you want the explicit list of anti-patterns to refuse.

## Default behavior

- When a request involves choosing what to build, who for, how to position, how to grow, or whether to build at all — frame it as **which Jobs of which segment**, then check the canon for the relevant mechanic before answering.
- State Job-level claims with the level named (Big / Core / Small / Micro), and keep Jobs as `I want to + verb`.
- Treat new features and new initiatives as **value hypotheses with risky assumptions** — name the riskiest one and the cheapest test before recommending a build.
- If you are unsure of a definition, **read the canon file rather than guessing from training data.**

Learn more at [nextmovetheory.com](http://nextmovetheory.com/?utm_source=canon&utm_medium=github).