How to Read the Equations

Why This Guide Exists

The equations in The Multiplier and the Mirror use notation drawn from calculus, statistics, and production economics. If you last encountered these symbols in a college course you have since forgotten, or if you never took that course at all, the notation can feel like a wall between you and the argument.

The wall is thinner than it looks. The paper uses approximately 40 distinct symbols and conventions. Most map to concepts you already reason about in everyday life: rates of change, averages, products, thresholds. This guide gives you the vocabulary to read every equation in the paper without needing to recall (or learn) the underlying mathematics from scratch.

The Appendix of the main paper provides “In plain language” glosses for each equation. This guide complements those glosses by covering the symbols themselves, the recurring building blocks that appear across many equations, so that you can read any equation on sight rather than looking up each one individually.

1. The Symbol Vocabulary

Mathematical notation is a language. Like any language, fluency comes from vocabulary. The vocabulary here is small: roughly 40 symbols, most of which map to something you already understand.

1.1 The Framework’s Own Variables

These are the paper’s named quantities. They are not standard mathematical symbols; they are defined by the framework.

1.2 Greek Letters

Greek letters are used as coefficients (explained in Section 2 below) and parameters: fixed or slowly changing numbers that control how fast things grow, shrink, or transfer. Think of them as the “settings” or “dials” on the system.

1.3 Mathematical Operators and Notation

These are the standard mathematical symbols that appear throughout. Each one has a plain-English reading.

1.4 Calculus Notation

These symbols describe rates of change. They are the framework’s way of talking about speed and acceleration.

1.5 Statistical Notation

1.6 Exponents, Subscripts, and Superscripts

1.7 Special Conventions in This Paper

2. What Is a Coefficient?

A coefficient is a fixed number that sits in front of a variable and controls how strongly that variable contributes to the result. It is the “dial setting” on a term in an equation.

2.1 The Concept

Consider a simple example from everyday life. If you earn $50 per hour, the equation for your daily earnings is:

\text{Earnings} = 50 \times \text{hours worked}

The number 50 is a coefficient. It does not change hour to hour. It tells you how much each hour contributes to the total. If you got a raise to $60/hour, the coefficient changes, and every hour now contributes more.

In the framework’s equations, the Greek letters \alpha, \beta, \gamma, \sigma, \delta, \phi, \psi, \eta, \lambda, \mu, \kappa, and \rho all serve as coefficients. They are the “dial settings” that determine how strongly each dynamic operates.

2.2 Why Coefficients Matter

Consider Eq. 11, the central dynamics equation:

\frac{dF}{dt} = \alpha \cdot S + \gamma \cdot E \cdot F - \beta \cdot R - \sigma \cdot M_{\text{absorbed}}

This equation has four terms, each controlled by its own coefficient:

The signs matter. The first two terms are positive (they add to dF/dt, meaning they grow Force). The last two are negative (they subtract from dF/dt, meaning they erode Force). The coefficients determine how strong each pressure is, and the balance between them determines whether Force grows or decays.

2.3 Coefficients vs. Variables

The distinction is important:

• Variables change during the analysis. F(t), M(t), S(t), R(t) all evolve over time. They are the things you are tracking.
• Coefficients are treated as fixed for a given context. \alpha, \beta, \gamma do not change moment to moment. They characterize the environment: how effective learning is, how corrosive passive reliance is, how responsive the organization is. They might differ between organizations, between individuals, or between eras, but within a given analysis they are constants.

This is why the tipping point equation (Eq. 14) is so revealing:

F^* = \frac{\beta \cdot R + \sigma \cdot M_{\text{absorbed}}}{\gamma \cdot E}

The tipping point is entirely determined by the coefficients and the behavioral variables (R, E, M_{\text{absorbed}}). It tells you: given this learning environment (these coefficient values) and this pattern of use (these behavioral values), here is the threshold. Change the coefficients (build a better learning culture, reduce passive reliance) and the threshold moves.

2.4 Coefficients with Subscripts

Some coefficients carry subscripts to indicate that the same type of dial operates differently in different contexts:

The subscript tells you where the dial applies. The decay rate for the surface layer (\beta_s) is much larger than the decay rate for the deep layer (\beta_d), which is written as \beta_d \ll \beta_s: surface skills rust quickly; deep intuition erodes slowly.

3. How to Read the Equation Types

The paper uses six main equation forms. Once you recognize the form, reading the content becomes straightforward.

3.1 Production Functions: “Output equals…”

What a production function is. A production function is an equation that says how much output a system produces, given its inputs. It is the mathematical language economists use for “this much labor, this much capital, these raw materials yield this much product.” The inputs appear on the right, the output on the left, and the function’s exact form describes how the inputs combine. In this framework the primary inputs are the multiplier M (the tool’s amplification) and Force F (the human capability being amplified), and the output O is productive work.

The content of a production function lives in its combining rule. An additive production function says more of one input compensates for less of another: strength in one dimension covers weakness in the next. A multiplicative production function says all inputs must be present at meaningful levels because any one near zero collapses the whole result. The paper’s central claim about Force is that it is multiplicative within a domain, which is why a single weak component (say, poor calibration of uncertainty) can drag the whole capability stack down.

Form: O = M \times F

How to read it: “Output equals the multiplier times Force.” This is the simplest form: one quantity is the product of two others. If either factor doubles, output doubles. If either goes to zero, output goes to zero.

The Cobb-Douglas extension (Eq. 1) expands F into a product of components. Cobb-Douglas is the name economists give to a specific family of multiplicative production functions in which each input is raised to a fixed exponent before being multiplied. The exponents control how sensitive the output is to changes in each input. The form is:

F = \prod_i f_i^{w_i}

Read the \prod as “multiply together”: take each component f_i, raise it to the power w_i, then multiply all the results. The key insight is that this is a product, not a sum. In a sum, a zero term is merely absent. In a product, a zero term kills the entire result, just as a single broken link breaks a chain regardless of how strong the other links are.

Pattern to recognize: When you see X = A \times B \times C, you are looking at a system where all factors must be present. Any factor near zero collapses the whole thing.

3.2 Differential Equations: “The rate of change is…”

What a differential equation is. An ordinary algebraic equation relates quantities: 3 + 4 = 7, or O = M \times F. A differential equation relates a quantity to its own rate of change. Its subject is not the current value of something but the speed at which that something is moving. The unknown in a differential equation is a function (usually a function of time), and the equation specifies how that function’s slope depends on the function’s current value and possibly on other variables or the clock.

Written generically, the simplest differential equation has the form

\frac{dy}{dt} = h(y, t)

read as “the rate at which y is changing at time t equals some formula h involving y and t themselves.” The derivative on the left is the rate; the expression on the right tells you what that rate depends on. If you know y right now, the equation tells you which direction y is moving and how fast.

Why this matters for the framework. Many claims in the paper are not about the current value of something but about how it evolves. Force grows from struggle and engagement; Force decays from reliance and organizational de-investment. Those are statements about rates, and the natural mathematical container for them is the differential equation. A handful of differential equations (Eqs. 11, 12, 23, 25, 33) carry most of the paper’s dynamic content.

Solving versus reading. To solve a differential equation is to find the function y(t) whose rate matches what the equation specifies: given the rate at every moment, recover the trajectory. This paper rarely solves its differential equations. It reads them. The structure alone (which terms appear, with what signs, in what combinations) is enough to answer qualitative questions such as: is there a steady state, does it repel or attract nearby trajectories, does the system bifurcate into multiple regimes? The Phase Portrait section (Eqs. 33 to 36) is an extended exercise in exactly this kind of reading.

Form: \dfrac{dF}{dt} = \text{(things that increase F)} - \text{(things that decrease F)}

How to read it: “The speed at which Force is changing equals the sum of all growth pressures minus all decay pressures.” Think of it as a bathtub: dF/dt is whether the water level is rising or falling. The positive terms are faucets (inflow). The negative terms are drains (outflow). If faucets exceed drains, the level rises. If drains exceed faucets, it falls. If they exactly balance, the level holds steady, and that steady level is the equilibrium (F^*).

The critical feature of Eq. 11 is that one of the growth terms, \gamma \cdot E \cdot F, contains F itself. This makes it a compounding term: the more Force you have, the faster it grows, like compound interest. This is what creates the tipping point. Below a certain level of F, the compounding term is too small to overcome the drains. Above it, the compounding term dominates and growth accelerates.

Second derivatives (d^2F/dt^2, Eq. 16a) tell you whether the rate of change is itself increasing or decreasing. If dF/dt is the speedometer, d^2F/dt^2 is whether you are pressing the accelerator or the brake.

Partial derivatives (\partial V / \partial I, Eq. 3) isolate one input. “If I change only the provider’s investment, holding everything else constant, how does skill value change?” It is like turning one dial on a mixing board while leaving all others fixed.

3.3 Statistical Equations: “The spread is…”

What a statistical equation is. A statistical equation describes a property of a whole distribution rather than of a single value. Where a production function asks “given these inputs, what output?”, a statistical equation asks “given this group, how spread out are they, how clumped around the mean, how correlated with each other?” The inputs and outputs are properties of populations: averages, variances, probabilities, correlations. The framework uses statistical equations to answer questions that cannot be posed about a single engineer, such as “if every engineer’s output is amplified by the same multiplier, what happens to the gap between the best and the worst?”

What variance is. The variance of a quantity is a measurement of how spread out its values are around their mean. If a group has values X_1, X_2, \ldots, X_n with mean \bar X, the variance is

\text{Var}(X) = \frac{1}{n} \sum_{i=1}^{n} (X_i - \bar X)^2

the average of the squared distances from each value to the mean. A distribution in which every value equals the mean has zero variance: no spread. A distribution in which values range widely from the mean has high variance. Variance is always non-negative because it is built from squared numbers, and it is expressed in squared units (if X is in dollars, \text{Var}(X) is in squared dollars).

The squared deviation is the standard choice because it penalizes large outliers disproportionately and admits cleaner algebra than the absolute deviation. One consequence of that choice, central to the framework: when you multiply every value in a distribution by a constant c, the variance does not scale by c; it scales by c^2. That scaling law is what Eq. 4 exploits.

Form: \text{Var}(O) = M^2 \times \text{Var}(F)

How to read it: “The spread in output equals the multiplier squared times the spread in Force.” This is a standard statistical property: when you multiply a distribution by a constant, the variance (the spread) scales by the square of that constant, not the constant itself.

Why the square? Variance measures squared deviations from the mean. Multiplying every value by M multiplies every deviation by M, and squaring a deviation that has been multiplied by M gives M^2 times the original squared deviation.

What this means practically: If your team has a moderate spread in capability and you introduce a 3x multiplier, the spread in output does not triple. It increases ninefold (3^2 = 9). A 5x multiplier produces a twenty-fivefold increase in spread. The multiplier is a divergence engine.

3.4 Threshold and Piecewise Equations: “It depends on which side you’re on”

What a piecewise function is. A piecewise function is a function that uses different formulas for different ranges of its input. Rather than a single rule that applies everywhere, the function is split into pieces, each governed by its own rule, with conditions that say which rule applies when. The conditions are mutually exclusive and together cover every possible input, so for any input exactly one rule fires. A piecewise function is the mathematical way of writing “if-then-else” logic inside an equation. The curly-brace notation \begin{cases} \ldots \end{cases} is read from top to bottom: check the first condition; if it holds, use its rule; otherwise check the next condition; and so on.

What a threshold is. A threshold is a critical value of some variable that separates regions of qualitatively different behavior. Crossing a threshold does not merely mean the variable is now larger or smaller; it means the system is in a different regime, governed by different rules. Freezing at 0°C, boiling at 100°C, and the threshold of pain in a stressed tissue are everyday examples: the underlying quantity varies continuously but its consequences switch discretely at the threshold. Thresholds often appear as the boundaries between the pieces of a piecewise function, and they are the values at which behavior bifurcates.

Form: V(F) \to \begin{cases} \text{Rule A} & \text{if condition 1} \\ \text{Rule B} & \text{if condition 2} \\ \text{Rule C} & \text{if condition 3} \end{cases}

How to read it: “Different formulas apply depending on which condition is met.” This is an if/else structure. The conditions are mutually exclusive; exactly one applies for any given input. In Eq. 6, the three conditions sort people into three labor-market tiers based on their Force level relative to a threshold.

The tipping point (Eq. 14) is a special case: a single value F^* that divides all possible trajectories into two basins. Above F^*, one set of dynamics dominates (growth compounds). Below F^*, the opposite set dominates (decay compounds). The equation itself is derived by finding the balance point where growth and decay exactly cancel, then recognizing that any displacement from that balance point pushes you further away from it rather than back toward it. This is what makes F^* a tipping point rather than a resting point.

3.5 Exponential Functions: “Compound growth or compound decay”

What an exponential function is. An exponential function is a function in which the variable appears in the exponent of a fixed base, rather than in the base itself. The generic form is f(t) = a \cdot b^{k t}, where a sets the starting value, b is the base, and k scales the rate. The defining property of an exponential function is that its rate of change is always proportional to its current value. Double the current value and the rate of change doubles. Halve it and the rate halves. Nothing else in elementary mathematics has that property, which is why exponentials are the natural language for compound growth and compound decay. Anything whose speed of change depends on how much of it is currently present (a population, a bank balance earning interest, an amount of radioactive material) is described by an exponential function.

What e is. The number e \approx 2.71828 is a fundamental mathematical constant, in the same family as \pi \approx 3.14159. It is the unique base for which the exponential function’s rate of change equals its current value exactly, not just proportionally. That is, e^t is the function whose derivative is itself: the slope at every point equals the height at that point. Any continuous-time growth or decay process with a rate proportional to its current size has a natural description in the form e^{kt}, where k is the rate of continuous compounding. The framework uses e for the same reason biology and finance use it: continuous-time compounding, not discrete snapshots, is the underlying model.

Two exponential forms recur throughout:

Growth: M(t) = M_0 \cdot e^{\mu t}

Read: “The multiplier at time t equals its starting value times e raised to the growth rate times time.” This is compound growth. The value at each moment grows in proportion to its current size, like money earning interest on interest. The larger it gets, the faster it grows in absolute terms. M_0 is where it starts; \mu controls how fast.

Decay: f_{\text{mot}}(t) = f_{\text{mot}}(0) \cdot e^{-\lambda \cdot A(t)}

Read: “Motivation at time t equals initial motivation times e raised to negative-lambda-times-accumulated-autonomy-loss.” The negative sign in the exponent makes this shrink over time instead of grow. Like a radioactive substance losing half its mass at regular intervals, the quantity falls toward zero but never quite reaches it. The rate of fall (\lambda) and the accumulated exposure (A(t)) together determine how much is left.

The key intuition: e^{+\text{something}} curves upward, growing faster and faster. e^{-\text{something}} curves downward, shrinking but never reaching zero. Both appear throughout the framework because the dynamics it describes are compounding: growth compounds, decay compounds, and which one dominates depends on the tipping point.

3.6 Coupled Dynamics and the Phase Plane

The last form of equation in the paper appears in the Phase Portrait section. Up to Eq. 32, Force dynamics (Eq. 11) and multiplier dynamics (Eq. 25) were written separately, with one treated as a backdrop while the other evolves. Eqs. 33 through 36 write both together. The mathematical object becomes a coupled system, and a new vocabulary is needed to read it.

Form:

\dot{F} = f(F, M), \qquad \dot{M} = g(F, M)

How to read it. “Force’s rate of change depends on both Force and the multiplier; the multiplier’s rate of change also depends on both.” Neither variable evolves on its own. Each is an input to the other’s rate. The state of the system at any moment is a pair (F, M), a single point on a two-dimensional plane. As time passes, the point moves, tracing a curve called a trajectory. The whole plane, with every possible starting point and the curve that emerges from it, is the phase portrait.

Monotone functions and monotonicity. A function is monotone if its direction of change is always the same. A monotonically increasing function never decreases; a monotonically decreasing function never increases. In derivative terms, f(x) is monotonically increasing if df/dx \geq 0 everywhere, and monotonically decreasing if df/dx \leq 0 everywhere. The word “monotonicity” names this property. A function can be monotone without being smooth, linear, or symmetric; the commitment is purely about the sign of change, not about shape or rate.

The Phase Portrait section uses monotonicity to describe the functions that appear in the coupled system without pinning down exact formulas. When the paper writes

\frac{\partial R}{\partial M} > 0, \qquad \frac{\partial S}{\partial M} < 0, \qquad \frac{\partial E}{\partial F} \geq 0, \qquad \mu'(\bar F) \geq 0

it is asserting only that passive reliance rises with the multiplier, productive struggle shrinks with it, deliberate engagement does not decrease as Force grows, and multiplier growth does not fall as workforce Force rises. The derivations rely only on those signs. Whether R grows linearly with M, logarithmically, or with an inflection point somewhere, the separatrix still rises and the basin classification still holds.

This is why the paper’s five candidate shapes of \mu(\bar F) (decoupled, linear, threshold-like, saturating, signed) are distinct empirical claims rather than distinct theoretical models. Each is a different way the same monotonicity commitment could be realized in the world. The framework does not choose between them; the choice is an empirical matter. The Phase Portrait’s structural conclusions, such as whether three distinct basins exist, depend on which shape turns out to describe reality.

Fixed points. A fixed point (or equilibrium) is a state (F^*, M^*) at which both rates equal zero: \dot F = 0 and \dot M = 0 simultaneously. A trajectory that arrives at a fixed point stops there, absent perturbation. In the phase portrait it is a single point; in practice it is a possible destination or a possible crossing.

Linearization and the Jacobian. The question “what happens if the system is nudged slightly away from a fixed point?” is answered by linearization: zoom in until the rates look linear, and study the linear system instead. For a coupled system with two rate equations, the linear approximation is captured in a single 2x2 matrix called the Jacobian.

Definition. For the system

\dot F = f(F, M), \qquad \dot M = g(F, M),

the Jacobian matrix evaluated at a point (F^*, M^*) is

J(F^*, M^*) = \begin{bmatrix} \dfrac{\partial f}{\partial F} & \dfrac{\partial f}{\partial M} \\[0.6em] \dfrac{\partial g}{\partial F} & \dfrac{\partial g}{\partial M} \end{bmatrix}_{(F^*, M^*)}.

Each entry is a partial derivative (see §1.4): the change in one rate caused by a small nudge in one variable, with the other variable held fixed. The subscript (F^*, M^*) means every entry is evaluated at the fixed point, so the four entries are ordinary numbers once the system and the point are specified. The row tells you which rate is responding; the column tells you which variable was nudged.

Reading the four entries in the Phase Portrait system. With f(F, M) = \dot F and g(F, M) = \dot M from Eq. 33:

Under the paper’s monotonicity commitments, the first column is usually positive (Force and the multiplier are self-reinforcing within range) and the off-diagonals have signs determined by the sign conditions on R(M), S(M), and \mu'(\bar F). Nothing about the exact magnitudes is fixed; they depend on functional forms the framework leaves empirical.

How the matrix acts. A small perturbation from the fixed point is a vector \delta = (\delta_F, \delta_M), describing how far off from (F^*, M^*) the system currently is. The Jacobian turns that perturbation into a rate vector. Multiplying the matrix by the vector gives

J \, \delta = \begin{bmatrix} \dfrac{\partial f}{\partial F} \delta_F + \dfrac{\partial f}{\partial M} \delta_M \\[0.6em] \dfrac{\partial g}{\partial F} \delta_F + \dfrac{\partial g}{\partial M} \delta_M \end{bmatrix}

which is approximately (\dot F, \dot M) for the nudged system. Read the top entry in plain language: “the rate at which F is moving is built from two pieces, one proportional to how far F is off, the other proportional to how far M is off, each weighted by how responsive \dot F is to that variable.” The bottom entry has the same structure for \dot M. The entire local dynamics near the fixed point is packaged in this one matrix-vector product.

Why linearization works. Any smooth rate function can be expanded around a fixed point as a Taylor series. The zeroth-order term vanishes because both rates are zero at a fixed point by definition. The first-order term is exactly J times the perturbation. Higher-order terms are second-order in the perturbation and become negligible as the perturbation shrinks. So for a sufficiently small neighborhood of the fixed point, studying \dot{\delta} \approx J \delta gives trustworthy answers about the real nonlinear system. That is why local stability questions can be answered by a matrix of constants rather than by solving the full nonlinear equations.

What an eigenvalue is. A matrix is a machine that takes in arrows and returns arrows. Push a vector through a 2x2 matrix and you generally get back a different vector, pointing in a new direction and with a new length. The matrix has rotated it and rescaled it at once. But every matrix has a small number of special directions in which no rotation happens: an input vector pointing along such a direction comes out still pointing along that direction, just stretched, shrunk, or flipped. Those special directions are called eigenvectors, and the amount of stretching along each one (including the sign) is the corresponding eigenvalue. A 2x2 matrix has, generically, two eigenvectors and two eigenvalues, one for each.

The word “eigen” is German for “own” or “characteristic.” An eigenvector is the matrix’s own preferred direction; the eigenvalue is the amount of stretching the matrix contributes along it. Together, an eigenvalue and its eigenvector describe a mode of behavior that the matrix leaves simple: pure stretch, no bend.

Why eigenvalues determine stability. Near a fixed point, the flow in the phase plane is approximately linear, with the linear part given by the Jacobian matrix J. Along an eigenvector direction of J, the dynamics reduce to a one-dimensional problem: a perturbation pointing that way simply gets stretched at a rate given by the corresponding eigenvalue. Concretely, a perturbation of size \delta along an eigenvector with eigenvalue \lambda evolves as \delta \cdot e^{\lambda t}.

This is the heart of the matter:

• If \lambda > 0, the perturbation grows exponentially. That direction is unstable.
• If \lambda < 0, the perturbation shrinks exponentially. That direction is stable.
• If \lambda is complex, say \lambda = a + b i, the real part a gives the growth-or-shrink rate and the imaginary part b gives a rotation rate: perturbations spiral. Positive a means outward spiral; negative a means inward spiral. Complex eigenvalues always come in conjugate pairs a + bi and a - bi, so the two principal modes are mirrored, and the trajectory spirals rather than slides.

A generic perturbation is a mixture of the two modes. Over time, the mode with the larger real-part eigenvalue comes to dominate and determines the trajectory’s long-run shape. That is why a single positive eigenvalue is enough to make a fixed point unstable even when the other eigenvalue is strongly negative: the positive mode eventually wins.

A concrete example. Suppose J = \begin{pmatrix} 2 & 0 \\ 0 & -1 \end{pmatrix}. Because this matrix is diagonal, its eigenvectors are the coordinate axes themselves, and its eigenvalues are the diagonal entries: \lambda_1 = 2 along the horizontal axis, \lambda_2 = -1 along the vertical. A perturbation in the horizontal direction grows as e^{2t}; a perturbation in the vertical direction shrinks as e^{-t}. A generic perturbation is pulled outward along the horizontal and pushed inward along the vertical. That is the saddle pattern: one positive eigenvalue, one negative.

For a matrix whose eigenvectors are not the coordinate axes, the logic is identical: two special directions still exist, just tilted. Finding them is a standard calculation, but for a 2x2 Jacobian the trace-determinant shortcut below lets you read off the classification without ever writing the eigenvectors down.

The four classifications. Given this reading of eigenvalues, the four qualitatively distinct possibilities for a 2x2 fixed point are:

The saddle case is the one that matters most for the Phase Portrait reading. It is the difference between “managed decline is a destination” (stable node) and “managed decline is a transit point” (saddle). The classification is not a matter of interpretation; it is a matter of the eigenvalues.

Reading the trace-determinant shortcut. For a 2x2 Jacobian, the two eigenvalues are the roots of

\lambda^2 - \text{tr}(J)\,\lambda + \det(J) = 0

which gives

\lambda = \frac{1}{2}\bigl(\text{tr}(J) \pm \sqrt{\text{tr}(J)^2 - 4\det(J)}\bigr).

You can read the classification from the signs of trace and determinant alone without solving for the eigenvalues explicitly:

• \det(J) < 0: saddle. (One eigenvalue is positive, one is negative, regardless of trace.)
• \det(J) > 0 and \text{tr}(J) < 0: stable (node if \text{tr}(J)^2 > 4\det(J), spiral otherwise).
• \det(J) > 0 and \text{tr}(J) > 0: unstable (same split between node and spiral).
• \det(J) = 0 or \text{tr}(J) = 0: boundary case; higher-order analysis needed.

This is the entire vocabulary you need to read Eq. 35.

The separatrix. In systems with more than one fixed point, the plane divides into basins of attraction: regions from which trajectories flow to one fixed point rather than another. The curve that separates two basins is called the separatrix. In Eq. 34, F^*(M) is the separatrix between the virtuous basin and the decay basin. The scalar F^* of Eq. 14 is the intersection of this curve with a specific horizontal line; extending the argument to an evolving multiplier produces a curve rather than a number.

What the phase plane adds to the framework. Earlier equations in the paper describe dynamics along a single axis: Force grows or decays. The phase plane admits that the axis itself is moving. A trajectory can cross from one basin to another without any change in its own Force, simply because the separatrix rose past it. That is the mathematical content of “the threshold moves.”

Pattern to recognize. When you see a pair of rate equations \dot F = f(F, M) and \dot M = g(F, M), you are looking at a phase-plane problem. The question to ask is no longer “which direction does F move?” but “where is the nearest fixed point, what is its Jacobian, what do its eigenvalues say, and which basin is the current (F, M) in?”

4. Workshop

The companion document F* Equations Workshop walks through every equation in the framework with concrete numbers. It uses a recurring cast of three engineers (Ada, Ben, Cora) to show how the same equation produces radically different results depending on the inputs. Each of the 36 equations (plus sub-equations) gets a worked numerical example with step-by-step arithmetic, comparison scenarios, and plain-language interpretation of the results.

Start there once you are comfortable with the symbol vocabulary (Sections 1-3 above).

5. Equation Dependency Map

The equations build on each other. This map shows which ones you need to understand before moving to the next.

Layer 1: Foundations (read these first)

• Eq. 1 (Base model: O = M \times F) establishes the core relationship
• Eq. 1a (Substitution by layer) explains why the LLM helps differently at different depths

Layer 2: Structure (builds on Layer 1)

• Eq. 2 (Variable multiplier by domain) decomposes M across domains
• Eq. 8 (Additive Force, negative components) introduces the alternative model where Force can go negative

Layer 3: Consequences (builds on Layers 1-2)

• Eqs. 4, 4a, 5 (Variance amplification) follow directly from Eq. 1
• Eq. 6 (Barbell effect) follows from Eqs. 4-5 and 1a
• Eq. 7, 7a (Evaluation bottleneck) follow from the creation-cost collapse implied by high M
• Eq. 9 (Damage) follows from Eq. 8
• Eq. 10 (Epistemic gap) introduces the M_p / M_s distinction from Mirror

Layer 4: Dynamics (builds on Layers 1-3)

• Eq. 11 (Force dynamics) is the central engine; depends on Eq. 1 for the meaning of F
• Eqs. 11a-c (Layered decay) decompose Eq. 11 by Force layer, using Eq. 1a
• Eqs. 12, 12a, 12b (Tacit knowledge) build a parallel dynamics model for organizations
• Eq. 13 (Pipeline break) is the threshold condition from Eq. 12

Layer 5: The Tipping Point (builds on Layer 4)

• Eq. 14 (Tipping point F^*) is derived from setting Eq. 11 to zero and solving
• Eq. 14a (Hysteresis) modifies Eq. 14 with asymmetric recovery
• Eqs. 15a, 15b (Diverging trajectories) follow from Eqs. 11 and 14
• Eqs. 16, 16a (Accelerating gap) follow from Eqs. 15a and 15b

Layer 6: Organizational and Systemic (builds on Layers 1-5)

• Eqs. 17-21 (ROI paradox, legibility crisis, Goodhart, decision bottleneck) apply the framework to organizational contexts
• Eq. 22 (Competitive moats) extends to inter-firm competition
• Eq. 23 (Motivation decay) feeds back into Eq. 1 and Eq. 11

Layer 7: Transfer and Sovereignty (builds on everything)

• Eqs. 24, 24a (Sovereignty) extend Eq. 1 to nations
• Eq. 25 (M growth) makes the multiplier dynamic
• Eqs. 26-31 (F-to-M transfer) close the loop between F and M
• Eq. 32 (Cohort discontinuity) introduces generational dynamics

Layer 8: The Phase Portrait (builds on Layers 4, 5, 7)

• Eqs. 33a, 33b (Coupled (F, M) system) rewrite Eqs. 11 and 25 as a simultaneous pair in which each variable drives the other
• Eq. 34 (Separatrix) generalizes the scalar F^* of Eq. 14 into a curve F^*(M) that shifts as M grows; Eq. 30 is recovered as a corollary
• Eq. 35 (Jacobian) is the local stability apparatus: four partial derivatives arranged in a 2x2 matrix whose eigenvalues classify each fixed point as stable, unstable, or saddle
• Eq. 36 (\Omega_{\text{irreversible}}) adds a second geometric object, the region from which no feasible policy recovers the trajectory; the frontier \partial \Omega gives hysteresis (Eq. 14a) a geometric expression

6. Reading Strategies

Start with the Base Model

Every equation in the paper is a descendant of O = M \times F (Eq. 1). If you understand that one equation, you have the seed from which everything else grows. Each subsequent equation asks: “What happens to this simple relationship when we relax an assumption?”

• “What if M varies by domain?” gives you Eq. 2.
• “What if F changes over time?” gives you Eq. 11.
• “What if F can go negative?” gives you Eq. 8.
• “What if there is a level of F where the dynamics flip?” gives you Eq. 14.

Read the “Captures” Line First

Each equation in the Appendix ends with a “Captures:” summary. Read that line before reading the equation itself. It tells you what the equation is saying in one sentence, which orients you before you parse the symbols.

Follow the Signs

In any equation with plus and minus terms (especially Eq. 11), the signs tell you the story. Positive terms are forces pushing upward (growth, building, compounding). Negative terms are forces pushing downward (decay, erosion, loss). The equation is a tug-of-war, and the signs tell you who is pulling in which direction.

Watch for Multiplication vs. Addition

The paper is precise about when it uses products (\prod, \times) and when it uses sums (\sum, +). Products mean “all factors must be present; a zero kills the whole thing.” Sums mean “contributions are independent; strength in one area compensates for weakness in another.” The structural difference between Eq. 1 (multiplicative, within a domain) and Eq. 2 (additive, across domains) is not a mathematical detail. It is the core of the argument about how Force works.

Use the Worked Examples as Templates

When an equation confuses you, plug in concrete numbers. The worked examples in Section 4 above are templates. Pick values that are easy to compute with (0.5, 1, 2, 3) and trace through the equation by hand. The abstract symbols become concrete quickly.

7. Notation Glossary (Quick Reference)

For quick lookup during reading. Sorted alphabetically within categories.

Roman Letters

Greek Letters