BEFORE THE FIRST PROJECTION

A tablet is a peculiar object. From across the room it is a shape. At arm’s length it is an inscription. Under the hand it becomes a sequence of cuts: stroke after stroke, relation after relation. The same surface can be carried as an image, read as language, or reconstructed as procedure. That is how this Tablet asks to be read. A glyph is compression: a mark, equation, phrase, diagram, or image that can hold more than the eye receives at first glance. A structure is what survives when the glyph is opened: the derivation, transformation, count, symmetry, balance, update, or change of basis that another reader could reproduce.

Depth begins only after the structure has earned it. Depth is the philosophical, Hermetic, literary, or mythic resonance that becomes possible once we know what the mathematics actually says—and what it does not. These are not three competing truths. They are three distances from the same page. A glyph without structure can become decoration. Structure without depth can become bloodless calculation. Depth without provenance can become mysticism pretending to be evidence.

The Egyptian register belongs here as an atmosphere of inscription, measure, reckoning, calendar, memory, and ordered record. Thoth/Djehuty may stand at the edge of the page as a figure of writing and measure; Hermes Trismegistus belongs to a later Greco-Egyptian and Hermetic world. Neither is evidence that ancient Egypt possessed modern field theory, Fourier analysis, quantum mechanics, or hidden machines. The mythic layer may stand beside the mathematical one. It may not impersonate it.

The Mentat posture is equally simple: separate observation from inference; preserve provenance; estimate before proclaiming precision; expose assumptions; ask what would change the conclusion; compress only after the reasoning can be reconstructed. A beautiful model should survive an admirer who is willing to destroy it. That is the standard here. A ridiculous mnemonic may survive too. If the chicken helps you remember the invariant, keep the chicken. Just do not promote the chicken to a law of nature. Two quiet axes remain available whenever the boundary needs to be made explicit: KNOWLEDGE STATUS — ESTABLISHED / DERIVED / FICTIONAL. INTERPRETIVE ROLE — LITERAL / ANALOGICAL / MYTHIC.

They are independent. An established equation may be used analogically. A fictional mechanism may be literal inside a fictional world. A mythic image may preserve a question worth asking without becoming evidence for a physical claim.

RESEMBLANCE IS NOT AUTHORITY.

Now turn the object.

PROJECTION 0 — OBJECT

An object larger than any one view of it.

Imagine a solid object held in darkness while a lamp moves around it. Its shadow changes. From one angle a symmetry appears. From another, depth. From another, an apparent discontinuity that vanishes when the light moves again. No shadow is the object. A shadow can still be exact about what it preserves. Mathematics lives inside this problem. A system may be described by coordinates, by a graph, by a probability distribution, by a differential equation, by a spectrum of frequencies, by conserved quantities, by a state vector, by a geometry, or by a set of admissible actions. Sometimes two descriptions are equivalent. Sometimes one throws information away. Sometimes two forms only look alike.

The first question is therefore not Which formula is true? It is what happened to the object when we changed the description? A change of coordinates is not the same operation as a change of basis; a projection may discard information; an average may hide detail; a probability distribution replaces a particular state with weights over possible states; a global criterion can summarize local behavior; a multi-agent model changes who can act; and a change of scale can make some variables disappear while others emerge. After any such move, the Tablet asks the same question: what survived?

THE RECEIVER IS ALSO PART OF THE PROBLEM

A representation does not merely stand between an object and a page.

It stands between an object and a receiver.

That receiver may be a coordinate system, an instrument, a language, a model, an ear, an eye, or a mind deciding what deserves attention.

Agustín Rayo’s work in the philosophy of logic and language sharpens one side of this problem: the relation between our language and the world it represents is itself something to inspect. A formal description may reveal genuine structure without exhausting the thing described. A sentence may be intelligible without settling what is possible. A notation may extend what can be reasoned about while still possessing limits of expression.

Jeanette Andrews’ illusion-based practice sharpens the other side: before a representation is interpreted, something must first enter awareness. Attention is selective. A convincing experience can be engineered without changing the underlying object. What appears obvious to an observer may therefore tell us as much about the path of attention as about the world being observed.

Her In Plain Listen makes the point unusually concrete. Information describing a historical magical method is transformed into Morse-derived notation, then into musical score, then into sound and performance. The underlying information can persist across several representations while the experience of receiving it changes radically.

So add two questions to the Tablet’s first rotation:

What survived the change of representation?
What never entered the receiver in the first place?

The first is a question about structure.

The second is a question about access.

Confusing them is one of the oldest ways to mistake a view for the object.

Five coordinates recur as the object turns: STATE, what must be specified now; TRANSFORMATION, what changes the state or its description; REPRESENTATION, the coordinates, basis, variables, symbols, or model in which the object appears; INFORMATION, what can be measured, inferred, transmitted, compressed, or lost; and COUPLING, what can influence what. Three further lenses cut across them: INVARIANCE, what remains unchanged under a stated transformation; SCALE, what disappears or emerges when the level of description changes; and CONSTRAINT, what cannot vary freely. These are not eight compartments. They are coordinates and lenses on one higher-dimensional intellectual object.

Recurrence is seductive. When the same mathematical form appears in two distant subjects, it is tempting to conclude that the subjects are secretly the same. Sometimes there really is a deep equivalence. More often, a transformation preserves only part of the structure, allowing a mathematical technique to migrate while the physical meaning changes. And sometimes the resemblance goes no deeper than notation.

A serious correspondence therefore has to be reconstructible. What structure is shared? What transformation exposes it? What remains invariant? What assumptions make the mapping valid? Which quantities acquire new meanings? Where does the mapping stop working? The last question matters as much as the first, because a useful correspondence has a boundary. Without one, resemblance becomes suggestion rather than analysis.

A correspondence without a break condition is a mood.

$$\boxed{\textbf{BEAUTY PROPOSES. EVIDENCE DISPOSES.}}$$

The Tablet is not a doctrine of hidden sameness. It is a discipline of comparison strong enough to transfer structure without stealing ontology. We begin with the smallest possible object that can already change its meaning when turned: a count.

I — POSSIBILITY

Unopened doors are already a geometry.

Before a system can move, it must have somewhere to move. Before an event can be probable, it must first be possible. Before evidence can update belief, there must be hypotheses to update. The first mathematics is therefore not motion. It is choice.

The space itself deserves inspection. Agustín Rayo’s work on logical space presses on a useful boundary: deciding which possibilities our representation can distinguish is not cleanly separable from deciding how the world is. What a language can describe, what a formalism can discriminate, and what a theory counts as a genuine alternative are related questions, but they are not automatically the same question.

For the Tablet, the operational lesson is simple:

declare the space before assigning weights inside it.

If a possibility was excluded by the representation rather than by the world, no later probability calculation can repair the omission.

1. A coefficient is a count of paths

Consider

$$(1+x)^3.$$

Written as a product,

$$(1+x)(1+x)(1+x).$$

To obtain an $x^2$ term, choose $x$ from exactly two factors and $1$ from the remaining factor. There are three ways:

$$(x)(x)(1),\qquad (x)(1)(x),\qquad (1)(x)(x).$$

So

$$(1+x)^3 = 1+3x+3x^2+x^3.$$

The coefficient $3$ is not decoration left behind by multiplication. It is a count of paths through the product. For $n$ factors, the number of ways to select exactly $k$ of them is

$$\binom{n}{k} = \frac{n!}{k!(n-k)!}.$$

Hence

$$\boxed{ (1+x)^n = \sum_{k=0}^{n}\binom{n}{k}x^k }$$

and, more generally,

$$(a+b)^n = \sum_{k=0}^{n}\binom{n}{k}a^{n-k}b^k.$$

The algebraic question asks: What coefficient appears beside $x^k$? The combinatorial question asks: How many paths through the multiplication produce $x^k$? The answer is the same integer. The representation changed. The count survived. That is the first genuine rotation of the Tablet.

2. The same count appears as recursion

The binomial coefficients can be generated without multiplying a polynomial at all. They satisfy

$$\boxed{ \binom{n}{k} = \binom{n-1}{k-1} + \binom{n-1}{k} }$$

because a chosen set of $k$ objects either contains one designated object or it does not. If it does, choose the remaining $k-1$ from the other $n-1$. If it does not, choose all $k$ from those $n-1$. The same count now appears as a recursive rule. A few rows are enough to recognize Pascal’s triangle:

$$\begin{array}{ccccccc} &&&1\\ &&1&&1\\ &1&&2&&1\\ 1&&3&&3&&1 \end{array}$$

Now discard the magnitudes and keep only parity—odd or even. The exact coefficients collapse into a binary pattern. Repeating that reduction across larger rows produces the familiar Sierpiński structure. The fractal was not inserted from outside. It was latent in the recursive arithmetic and exposed by a new projection: integer magnitude became parity. This is a good place to practice restraint. The Sierpiński pattern is real. Its recurrence across scale is real. None of that proves that Pascal’s triangle is the secret causal engine of nature, consciousness, pyramids, markets, or anything else that happens to look triangular. The pattern may be a glyph. The rule is the structure. The ontology must still be earned.

3. A state space is possibility under distinctions and constraints

Suppose six positions can each occupy one of two states. The number of complete configurations is

$$2^6=64.$$

If we group those configurations by the number $r$ of positions in one designated state, then

$$\sum_{r=0}^{6}\binom{6}{r} = 1+6+15+20+15+6+1 = 64.$$

The same sixty-four possibilities have been reorganized by a new descriptive coordinate: how many positions carry one state. Permit three possible states at each position and the total becomes

$$3^6=729.$$

The system did not become mysterious. The alphabet changed. So did the possibility space. A model starts before the calculation. It starts when we decide what differences count. What is a state? Which distinctions are retained? Which are ignored? Which configurations are admissible? Which transitions are forbidden? A possibility space is not the world before we have measured it. It is the world under a chosen description and a set of constraints.

4. Possibility becomes probability when paths receive weight

Take a repeated binary experiment. Let one outcome have probability $p$ and the other $1-p$. Then

$$(p+(1-p))^n=1.$$

Expand:

$$1 = \sum_{k=0}^{n} \binom{n}{k} p^k(1-p)^{n-k}.$$

The same coefficient $\binom{n}{k}$ that counted algebraic paths now counts how many length-$n$ sequences contain exactly $k$ occurrences of one outcome. Weight each such sequence by $p^k(1-p)^{n-k}$ and we obtain

$$\boxed{ P(K=k) = \binom{n}{k}p^k(1-p)^{n-k} }$$

for the binomial model. Nothing was thrown away. Counting was weighted. Possibility became probability. The distinction will matter later. A state can be possible but extremely unlikely. A state can be likely but unobserved. An observation can be real but poorly measured. A measurement can be precise while the model interpreting it is wrong. Do not confuse existence, likelihood, observation, and evidence merely because all four can be written beside numbers.

5. A door left open

Return to the binomial, but change its scale:

$$\left(1+\frac{x}{n}\right)^n.$$

It still contains the same $\binom{n}{k}$. Now each individual change becomes smaller while the number of changes becomes larger. Finite counting is about to become continuous evolution. We do not have to declare the connection. We can watch it happen.

II — EVOLUTION

One step, repeated until the staircase becomes a curve.

Possibility tells us where a system may be. Evolution tells us how a state gets from here to there. The bridge between them is already sitting inside the binomial coefficients.

1. Normalized counting becomes the exponential

Start again:

$$\left(1+\frac{x}{n}\right)^n.$$

Expand before taking the limit:

$$\left(1+\frac{x}{n}\right)^n = \sum_{k=0}^{n} \binom{n}{k}\frac{x^k}{n^k}.$$

For a fixed $k$,

$$\binom{n}{k}\frac{1}{n^k} = \frac{n(n-1)\cdots(n-k+1)}{k!\,n^k}.$$

Factor out the powers of $n$:

$$\binom{n}{k}\frac{1}{n^k} = \frac{1}{k!} \left(1\right) \left(1-\frac1n\right) \left(1-\frac2n\right) \cdots \left(1-\frac{k-1}{n}\right).$$

For fixed $k$, every factor in parentheses tends to $1$ as $n\to\infty$. Thus

$$\boxed{ \binom{n}{k}\frac{1}{n^k} \longrightarrow \frac{1}{k!} }$$

and the coefficients of the exponential series emerge from normalized combinatorial counts:

$$\boxed{ e^x = \sum_{k=0}^{\infty}\frac{x^k}{k!} }$$

or

$$e^x = 1+x+\frac{x^2}{2!}+\frac{x^3}{3!}+\cdots.$$

A rigorous proof of the whole limiting process requires more care than termwise inspection. But the structural event is already visible. The factorials did not fall from the sky. We watched them emerge from counting.

2. Repeated finite change becomes continuous growth

Now give the same limit a dynamical meaning. Suppose a quantity changes over a time interval $t$ by repeatedly applying a small fractional change. Divide the interval into $N$ steps of width

$$\Delta t=\frac{t}{N}.$$

If the local growth rate is $\lambda$, use the update

$$x_{m+1}=(1+\lambda\Delta t)x_m.$$

After $N$ steps,

$$x_N = (1+\lambda\Delta t)^N x_0 = \left(1+\frac{\lambda t}{N}\right)^N x_0.$$

Let $N\to\infty$:

$$\boxed{ x(t)=e^{\lambda t}x(0). }$$

The finite update has become the solution of

$$\dot x=\lambda x.$$

The same limit that transformed binomial counting into the exponential has transformed repeated small updates into continuous evolution. That is not an analogy. It is the same limiting structure wearing a dynamical meaning.

3. The exponential reproduces itself under change

Differentiate its series:

$$\frac{d}{dx}e^x = \frac{d}{dx} \left( 1+x+\frac{x^2}{2!}+\frac{x^3}{3!}+\cdots \right).$$

Then

$$\frac{d}{dx}e^x = 1+x+\frac{x^2}{2!}+\frac{x^3}{3!}+\cdots = e^x.$$

Therefore

$$\boxed{\frac{d}{dx}e^x=e^x.}$$

The function reproduces its own shape under differentiation. Only a multiplier changes when the argument is scaled. That property is why exponentials appear so naturally in linear evolution. If $\lambda>0$, the mode grows. If $\lambda<0$, it decays. Soon $\lambda$ will become imaginary and growth will become rotation. But first there is another structural fact to earn.

4. Addition becomes multiplication

The exponential turns addition into composition:

$$\boxed{e^{x+y}=e^xe^y.}$$

The binomial theorem is still inside the proof. Begin with

$$e^{x+y} = \sum_{n=0}^{\infty}\frac{(x+y)^n}{n!}.$$

Expand the power:

$$(x+y)^n = \sum_{k=0}^{n}\binom{n}{k}x^k y^{n-k}.$$

So

$$e^{x+y} = \sum_{n=0}^{\infty} \sum_{k=0}^{n} \frac{1}{n!}\binom{n}{k}x^ky^{n-k}.$$

Because

$$\frac{1}{n!}\binom{n}{k} = \frac{1}{k!(n-k)!},$$

the terms factor into

$$\frac{x^k}{k!}\frac{y^{n-k}}{(n-k)!}.$$

Absolute convergence permits the regrouping, giving

$$e^{x+y} = \left(\sum_{k=0}^{\infty}\frac{x^k}{k!}\right) \left(\sum_{m=0}^{\infty}\frac{y^m}{m!}\right) = e^xe^y.$$

The bookkeeping of choice has become a law of composition. Add the generator. Multiply the evolution. That sentence will change meaning when the generator becomes a matrix, and again when it becomes a rotation.

5. One dimension becomes many

A scalar system has one state variable. A coupled system can have many. Write

$$\dot{\mathbf x}=A\mathbf x.$$

For a time-invariant matrix $A$, the solution is

$$\boxed{ \mathbf x(t)=e^{At}\mathbf x(0), }$$

where the matrix exponential is defined by the same series:

$$e^{At} = I+At+\frac{(At)^2}{2!}+\frac{(At)^3}{3!}+\cdots.$$

The scalar generator $\lambda$ has become a matrix generator $A$. Now suppose $A$ is diagonalizable:

$$A=V\Lambda V^{-1}.$$

Then

$$A^n=V\Lambda^nV^{-1},$$

so the entire exponential series transforms with it:

$$\begin{aligned} e^{At} &=I+At+\frac{A^2t^2}{2!}+\cdots\\ &=V\left(I+\Lambda t+\frac{\Lambda^2t^2}{2!}+\cdots\right)V^{-1}\\ &=Ve^{\Lambda t}V^{-1}. \end{aligned}$$

In the original coordinates, the components may be coupled. In the eigenvector basis, $\Lambda$ is diagonal and each mode evolves by its own scalar exponential. A difficult evolution has not been destroyed. It has been represented in coordinates that expose its independent modes. Now the question changes. Not merely: How does the state evolve? But: In what representation does the evolution become simple?

6. The generator is the code — but not one universal code

At this point a tempting mistake becomes possible. Because $e$ keeps returning, one might try to promote the number itself into a universal code of reality. That goes too far. The exponential is not a catalogue of all possible choices, and $\hbar$ is not a universal grid into which every event must fit. What survives across several theories is more structural:

$$\boxed{ \textbf{STATE + GENERATOR + CONSTRAINTS + INITIAL CONDITIONS} \;\longrightarrow\; \textbf{EVOLUTION} }$$

Within a specified model, the generator is the object that determines the local rule of change.

For one scalar degree of freedom,

$$\dot x=\lambda x$$

and

$$x(t)=e^{\lambda t}x(0).$$

For a time-invariant linear coupled system,

$$\dot{\mathbf x}=A\mathbf x$$

and

$$\mathbf x(t)=e^{At}\mathbf x(0).$$

Here $\lambda$ and $A$ are different kinds of generators. The exponential accumulates their infinitesimal rule into finite evolution. A useful compression is therefore:

$$\boxed{\textbf{THE GENERATOR WRITES THE LOCAL RULE. THE EVOLUTION UNFOLDS IT THROUGH TIME.}}$$

Classical Hamiltonian mechanics reaches the same architectural question by another route. A Hamiltonian $H(q,p,t)$ generates motion in phase space through Hamilton’s equations,

$$\boxed{ \dot q=\frac{\partial H}{\partial p}, \qquad \dot p=-\frac{\partial H}{\partial q}. }$$

For many familiar systems the Hamiltonian is the total energy, but its deeper role here is dynamical: together with the canonical phase-space structure, it determines how the state moves.

Quantum mechanics gives the word generator an especially clean form. The state obeys the Schrödinger equation

$$\boxed{ \mathrm{i}\hbar\frac{\partial}{\partial t}\lvert\psi(t)\rangle = \hat H\lvert\psi(t)\rangle. }$$

If the Hamiltonian is time-independent, then

$$\boxed{ \lvert\psi(t)\rangle = e^{-\mathrm{i}\hat Ht/\hbar}\lvert\psi(0)\rangle. }$$

Now the earlier pattern returns with a different physical meaning. The Hamiltonian $\hat H$ is the generator of time evolution; $\hbar$ sets the quantum scale; the exponential turns the generator into a finite evolution operator. The resemblance to $e^{At}$ is mathematically legitimate, but the objects are not interchangeable. A classical matrix, a classical Hamiltonian function, and a quantum Hamiltonian operator belong to different frameworks.

The simple exponential also has a break condition. If a quantum Hamiltonian varies with time and fails to commute with itself at different times, the evolution cannot in general be written as the naive $e^{-\mathrm{i}\hat Ht/\hbar}$; time ordering is required. The lesson is not that exponentials explain everything. It is that a generator-plus-flow architecture recurs, and its exact form depends on the structure being generated.

7. When the state is a probability distribution

So far the state itself has been a scalar, vector, phase-space point, or quantum state. Probability has appeared as a weight over possibilities and later will appear as a state of knowledge. There is another exact use: a probability distribution can itself be the evolving state.

Let

$$\mathbf p_n=(p_n(0),p_n(1),\ldots,p_n(m-1))$$

be a row-vector probability distribution on a finite state space. A transition matrix $P$ is row-stochastic when

$$P_{ij}\ge 0, \qquad \sum_jP_{ij}=1.$$

The next distribution is

$$\boxed{\mathbf p_{n+1}=\mathbf p_nP.}$$

The stochastic conditions are doing real work. If $\mathbf p_n$ has nonnegative entries that sum to one, then so does $\mathbf p_nP$ because

$$\sum_j(\mathbf p_nP)_j = \sum_i p_n(i)\sum_jP_{ij} = \sum_i p_n(i) =1.$$

Iterate the same transition rule:

$$\boxed{\mathbf p_n=\mathbf p_0P^n.}$$

This resembles the repeated application of a deterministic linear operator, but the admissible state space has changed. $P$ must preserve the probability simplex. Its entries are transition probabilities, not arbitrary couplings.

A stationary distribution is a probability row vector $\boldsymbol\pi$ that survives one transition unchanged:

$$\boxed{\boldsymbol\pi P=\boldsymbol\pi.}$$

That is immediately an eigenvector statement. $\boldsymbol\pi$ is a left eigenvector of $P$ with eigenvalue $1$. Stationarity has rotated into spectral structure.

The same idea has a continuous-time form. For a time-homogeneous finite-state process, write the short-time transition matrix as

$$P(\Delta t) = I+Q\Delta t+o(\Delta t),$$

where the generator $Q$ obeys

$$Q_{ij}\ge0\quad(i\ne j), \qquad \sum_jQ_{ij}=0.$$

The off-diagonal entries are transition rates; the diagonal balances the total rate leaving each state. Apply the short-time transition to the row distribution:

$$\mathbf p(t+\Delta t) = \mathbf p(t)P(\Delta t).$$

Subtract $\mathbf p(t)$, divide by $\Delta t$, and take the limit:

$$\boxed{\dot{\mathbf p}=\mathbf pQ.}$$

For time-independent $Q$,

$$\boxed{\mathbf p(t)=\mathbf p(0)e^{tQ}.}$$

The matrix exponential has returned, but with a new constraint: the generated flow must remain a probability distribution. A stationary distribution now satisfies

$$\boxed{\boldsymbol\pi Q=0.}$$

The stationary eigenvalue has shifted from $1$ for the discrete transition operator to $0$ for its continuous-time generator.

One last separation prevents two different uses of probability from collapsing into each other. BAYESIAN UPDATE: probability changes because evidence arrived. MARKOV EVOLUTION: probability changes because the modeled process evolved. Both move distributions. The transformation is different. So are the questions being answered.

So the Tablet can answer the informal question Where is the code? without pretending that physics has supplied one cosmic number:

The code is not a number. It is the rule that turns one admissible state into the next.

The next Projection asks what happens when a change of representation makes that rule simpler to see.

III — MODES & REPRESENTATION

The same star redrawn on another grid.

A difficult object can become simple without becoming different. That sentence is easy to admire and easy to misunderstand. We will earn it.

$$\boxed{\textbf{SAME OBJECT, DIFFERENT BASIS.}}$$

The exponential will now rotate. Literally.

1. Put $\mathrm{i}$ into the exponential

From the previous Projection,

$$e^x = 1+x+\frac{x^2}{2!}+\frac{x^3}{3!}+\cdots.$$

Replace $x$ with $\mathrm{i}\theta$:

$$e^{\mathrm{i}\theta} = 1+\mathrm{i}\theta +\frac{(\mathrm{i}\theta)^2}{2!} +\frac{(\mathrm{i}\theta)^3}{3!} +\cdots.$$

The powers of $i$ cycle:

$$1,\quad \mathrm{i},\quad -1,\quad -\mathrm{i},\quad 1,\ldots$$

Separate even and odd powers:

$$\begin{aligned} e^{\mathrm{i}\theta} &= \left( 1-\frac{\theta^2}{2!}+\frac{\theta^4}{4!}-\cdots \right)\\ &\quad+\mathrm{i}\left( \theta-\frac{\theta^3}{3!}+\frac{\theta^5}{5!}-\cdots \right). \end{aligned}$$

The first series is $\cos\theta$ and the second is $\sin\theta$. Therefore

$$\boxed{e^{\mathrm{i}\theta}=\cos\theta+\mathrm{i}\sin\theta.}$$

The exponential that described growth and decay now traces the unit circle. At $\theta=0$,

$$e^{\mathrm{i}0}=1.$$

At $\theta=\pi/2$,

$$e^{\mathrm{i}\pi/2}=\mathrm{i}.$$

At $\theta=\pi$,

$$e^{\mathrm{i}\pi}=-1.$$

At $\theta=2\pi$,

$$e^{\mathrm{i}2\pi}=1.$$

The state returns to the same visible point after one full turn. Already, a warning from the end of the book is hiding here: return of one coordinate does not guarantee return of every hidden state. For now, however, the unit circle is exact. Euler’s identity,

$$e^{\mathrm{i}\pi}+1=0,$$

is famous because it is beautiful. The more reusable fact is less ornamental:

$$\boxed{\text{exponential evolution can encode rotation}.}$$

The form did not change. The generator did.

2. Addition becomes composition of rotations

The exponential law survives:

$$e^{\mathrm{i}(\alpha+\beta)}=e^{\mathrm{i}\alpha}e^{\mathrm{i}\beta}.$$

Now addition of angles becomes multiplication of rotations. Use Euler’s formula on both sides:

$$\cos(\alpha+\beta)+\mathrm{i}\sin(\alpha+\beta) = (\cos\alpha+\mathrm{i}\sin\alpha)(\cos\beta+\mathrm{i}\sin\beta).$$

Multiply the right side:

$$\begin{aligned} &\cos\alpha\cos\beta-\sin\alpha\sin\beta\\ &\quad+\mathrm{i}\left(\sin\alpha\cos\beta+\cos\alpha\sin\beta\right). \end{aligned}$$

Match real and imaginary parts:

$$\boxed{ \cos(\alpha+\beta) = \cos\alpha\cos\beta-\sin\alpha\sin\beta }$$

and

$$\boxed{ \sin(\alpha+\beta) = \sin\alpha\cos\beta+\cos\alpha\sin\beta. }$$

The trigonometric addition formulas are inside the multiplicative law of the complex exponential. And the multiplicative law of the exponential was exposed through the same combinatorial structure that began with the binomial theorem. The chain is no longer rhetorical:

$$\binom{n}{k} \longrightarrow e^x \longrightarrow e^{x+y}=e^xe^y \longrightarrow e^{\mathrm{i}\theta} \longrightarrow \text{rotation}.$$

3. The matrix version of $i$

The same rotational structure can be written without complex numbers. Define

$$J= \begin{pmatrix} 0&-1\\ 1&0 \end{pmatrix}.$$

Then

$$J^2=-I.$$

So $J$ has the same algebraic square as $i$. Exponentiate:

$$e^{\theta J} = I+\theta J+\frac{\theta^2J^2}{2!}+\frac{\theta^3J^3}{3!}+\cdots.$$

Because $J^2=-I$, the even and odd powers split exactly as before:

$$e^{\theta J} = I\cos\theta+J\sin\theta.$$

Hence

$$\boxed{ e^{\theta J} = \begin{pmatrix} \cos\theta&-\sin\theta\\ \sin\theta&\cos\theta \end{pmatrix}. }$$

One representation uses multiplication by the complex number $e^{\mathrm{i}\theta}$. Another uses a real $2\times2$ matrix. The rotational action is the same. The symbols are not. This is the Tablet in miniature: a structural invariant revealed by changing representation.

4. Rotation through time becomes oscillation

Let the angle advance uniformly:

$$\theta=\omega t.$$

Then

$$e^{\mathrm{i}\omega t} = \cos(\omega t)+\mathrm{i}\sin(\omega t).$$

The real and imaginary coordinates of uniform circular motion are sinusoids. A sinusoid can therefore be understood as a projection of uniform rotation in a plane. Now watch the same structure appear from a differential equation. For a harmonic oscillator,

$$\ddot x+\omega^2x=0.$$

Try an exponential solution $x=e^{rt}$. Then

$$r^2e^{rt}+\omega^2e^{rt}=0,$$

so

$$r^2+\omega^2=0$$

and therefore

$$r=\pm \mathrm{i}\omega.$$

Thus

$$x(t)=c_1e^{\mathrm{i}\omega t}+c_2e^{-\mathrm{i}\omega t},$$

which, for real motion, is equivalent to

$$\boxed{ x(t)=A\cos(\omega t)+B\sin(\omega t). }$$

The imaginary roots are not decorative complex machinery pasted onto the oscillator after the fact. They are what the differential equation asks for. Rotation and oscillation have met from both directions. This does not mean every oscillating physical system contains a literal hidden wheel. The shared structure is exact at the level of the representation. The physical mechanism still belongs to the system being modeled.

5. When phases meet: superposition, interference, resonance

One rotating mode is simple. Put two together and phase becomes consequence.

For a linear operator $\mathcal{L}$, if

$$\mathcal{L}[u_1]=0, \qquad \mathcal{L}[u_2]=0,$$

then

$$\boxed{ \mathcal{L}[a u_1+b u_2]=0. }$$

This is the principle of superposition. It is enormously powerful and sharply bounded: it follows from linearity. Nonlinear systems do not, in general, preserve arbitrary sums of solutions.

Now let two equal-amplitude oscillations have the same angular frequency but a phase difference $\varphi$:

$$\cos(\omega t)+\cos(\omega t+\varphi) = 2\cos\left(\frac{\varphi}{2}\right) \cos\left(\omega t+\frac{\varphi}{2}\right).$$

The ingredients have not changed, but their relation has. When $\varphi=0$, they reinforce. When $\varphi=\pi$, they cancel. Between those limits the resultant amplitude depends continuously on phase. Interference is what superposed contributions look like when relative phase matters. No extra force has been invented; the sum itself has acquired structure.

A different question appears when a system is driven. Consider the damped, forced oscillator

$$\ddot x+2\gamma\dot x+\omega_0^2x = \frac{F_0}{m}\cos(\omega t).$$

Its steady-state amplitude is

$$\boxed{ A_{\mathrm{resp}}(\omega) = \frac{F_0/m} {\sqrt{(\omega_0^2-\omega^2)^2+(2\gamma\omega)^2}}. }$$

For weak damping, the response becomes large when the driving frequency lies near the system’s natural frequency. That is resonance. The word does not mean that matching frequencies always produce unlimited growth. Damping, coupling, forcing strength, boundary conditions, and the system’s own modes determine the response; with damping, even the precise peak can shift.

Now impose periodicity. If a signal has period $T$, define the fundamental angular frequency

$$\omega_0=\frac{2\pi}{T}.$$

Under the usual conditions for Fourier expansion, a periodic signal may be represented as

$$\boxed{ f(t)=\sum_{n=-\infty}^{\infty}c_n e^{\mathrm{i}n\omega_0 t}. }$$

Here the frequencies $n\omega_0$ are harmonics of the fundamental. This is a literal mathematical use of the word: a periodic object has been expressed as a weighted collection of rotating modes at integer multiples of one base frequency.

Acoustics and music can inhabit these structures, but music is not reducible to them. Tuning systems, temperament, timbre, psychoacoustics, rhythm, phrasing, culture, and artistic choice introduce additional structures and meanings. The Tablet keeps the transferable mathematics. The music corpus keeps the craft of making those relations audible.

Coupling will now make the idea less individual. When oscillators or state variables influence one another, their natural collective motions can become normal modes. The next turn asks which basis makes those coupled motions visible.

6. Why Fourier uses these rotations

Complex exponentials are unusually convenient because differentiation preserves their form:

$$\frac{d}{dt}e^{\mathrm{i}\omega t} = \mathrm{i}\omega e^{\mathrm{i}\omega t}.$$

Differentiate twice and only the multiplier changes again:

$$\frac{d^2}{dt^2}e^{\mathrm{i}\omega t} = -\omega^2e^{\mathrm{i}\omega t}.$$

A complicated operation in the original representation can become multiplication in the modal one. For one common Fourier-transform convention,

$$\hat f(\omega) = \int_{-\infty}^{\infty}f(t)e^{-\mathrm{i}\omega t}\,dt$$

and

$$\boxed{ f(t) = \frac{1}{2\pi} \int_{-\infty}^{\infty} \hat f(\omega)e^{\mathrm{i}\omega t}\,d\omega. }$$

In time, the signal may look tangled. In frequency, it becomes a weighted field of rotating modes. The representation has changed. The signal has not become more true. But some operations have become easier to see.

7. Coupled systems become modes when the basis cooperates

Return to

$$\dot{\mathbf x}=A\mathbf x.$$

If $A=V\Lambda V^{-1}$, define modal coordinates

$$\mathbf z=V^{-1}\mathbf x.$$

Then

$$\dot{\mathbf z} = V^{-1}\dot{\mathbf x} = V^{-1}AV\mathbf z = \Lambda\mathbf z.$$

If $\Lambda$ is diagonal,

$$\dot z_j=\lambda_j z_j$$

for each mode $j$, so

$$z_j(t)=e^{\lambda_jt}z_j(0).$$

In the original basis, the components may push and pull on one another. In the modal basis, each coordinate can evolve independently. Eigenvectors, normal modes, Fourier modes, and complex exponentials are not one object under different names. They repeatedly answer the same structural question: Which representation makes the transformation simple? That is a powerful question. It is not enough. Once we rotate the description, we must ask which features were real enough to survive the turn.

WORKED ROTATION — RETURN TO THE LAZY CYCLE

Projection II left us with a stochastic transition operator acting on a probability distribution. Now that Fourier representation and modal coordinates have been earned, return to that operator. Let $S$ be the cyclic shift on $N$ states and take the lazy walk

$$P = \frac12I+\frac14S+\frac14S^{-1}.$$

This particular $P$ is circulant. Circulant matrices are diagonalized by the discrete Fourier basis. With state index $j=0,\ldots,N-1$, take

$$v_k(j)=e^{2\pi\mathrm{i}kj/N}.$$

Then

$$Pv_k=\lambda_kv_k,$$

with

$$\boxed{ \lambda_k = \frac12+\frac12\cos\frac{2\pi k}{N}. }$$

If $\widehat p_n(k)$ denotes the Fourier coefficient of the probability distribution in this basis, repeated transition becomes mode-by-mode multiplication:

$$\boxed{ \widehat p_n(k) = \widehat p_0(k)\lambda_k^n. }$$

The $k=0$ mode has $\lambda_0=1$ and is constant around the cycle; normalized as a probability distribution, it is the uniform stationary mode. For every nonconstant Fourier mode of this lazy cycle walk, $0\le\lambda_k<1$, so repeated application of $P$ suppresses that mode. What looked like a random walk in the state basis has become independent scalar evolution in the Fourier basis:

$$\boxed{ \text{TRANSITION OPERATOR} \longrightarrow \text{EIGENBASIS} \longrightarrow \text{MODE-BY-MODE EVOLUTION} \longrightarrow \text{STATIONARITY}. }$$

The boundary matters. Ordinary discrete Fourier modes diagonalize this example because the transition matrix is circulant. A general Markov matrix can have a different eigenbasis, can be nonnormal, can fail to diagonalize cleanly, and need not inherit this simple Fourier picture. The exact rotation belongs to this structured example, not to every stochastic process.

A representation can make an operator transparent. It still need not be the thing an observer perceives.

8. The spectrum is not the percept

A Fourier representation can tell us which sinusoidal components are present in a signal. It does not, by itself, tell us what an auditory system will hear. One bounded example is enough to make the distinction visible.

Construct a harmonic complex with the first harmonic deliberately absent:

$$s(t) = \sum_{n=2}^{N} a_n\cos(2\pi n f_0t+\varphi_n).$$

By construction, this sum contains components at $2f_0,3f_0,\ldots,Nf_0$, but no Fourier component at $f_0$. Now shift time by $1/f_0$:

$$\begin{aligned} s\!\left(t+\frac1{f_0}\right) &= \sum_{n=2}^{N} a_n \cos\!\left(2\pi n f_0t+2\pi n+\varphi_n\right)\\ &= s(t). \end{aligned}$$

So $1/f_0$ is a period of this constructed signal even though the spectral line at $f_0$ is absent. It is not automatically the minimal fundamental period for every possible selection of nonzero amplitudes; a degenerate subset of active harmonic indices can repeat more frequently. The point is narrower and exact: periodic structure associated with $f_0$ can survive the removal of the $f_0$ spectral component.

The next step is empirical rather than algebraic. Experiments with missing-fundamental tones show that, under appropriate listening conditions, listeners may perceive or judge a pitch near the absent $f_0$ even though there is no physical spectral component there. Listener and stimulus differences matter; the percept is not forced by the equation alone.

That gives us a new chain, but the arrows do not all mean the same thing:

$$\boxed{ \text{PHYSICAL SIGNAL} \longrightarrow \text{SPECTRAL REPRESENTATION} \longrightarrow \text{SENSORY PROCESSING} \longrightarrow \text{PERCEPT}. }$$

The first turn can be an exact mathematical representation under the stated Fourier conditions. The later turns involve a biological sensory system and empirical perceptual inference. A perceived pitch near $f_0$ is therefore not evidence that a hidden oscillator at $f_0$ was physically added to the waveform.

$$\boxed{\textbf{AN EMERGENT PERCEPT DOES NOT IMPLY AN ADDITIONAL PHYSICAL SOURCE.}}$$

This distinction will matter in Projection V, where observation, model, and inference are separated explicitly. A percept is evidence available to an observer. It is not a transparent copy of every cause that produced it.

IV — SHAPE, SYMMETRY & INVARIANCE

The figure turns. Something refuses to move.

An invariant is never merely a thing that stays the same. It stays the same under something. Without the transformation, the word invariant is unfinished.

1. Invariance must name the transformation

Let a vector $\mathbf x$ be rotated by an orthogonal matrix $R$ satisfying

$$R^\top R=I.$$

Its squared length after rotation is

$$\|R\mathbf x\|^2 = (R\mathbf x)^\top(R\mathbf x).$$

Therefore

$$\begin{aligned} \|R\mathbf x\|^2 &= \mathbf x^\top R^\top R\mathbf x\\ &= \mathbf x^\top\mathbf x\\ &= \|\mathbf x\|^2. \end{aligned}$$

The coordinates changed while the length did not. That is the invariance statement: not that nothing changed, but that this quantity survived this transformation. Precision begins by naming both the transformation and the survivor.

2. Symmetry is a transformation that preserves specified structure

A square survives a rotation by $90^\circ$ as the same geometric set. A physical law may retain its form after translating the origin. A matrix may change every visible entry after a change of basis while preserving its eigenvalues. A topological object may be stretched and bent while preserving connectedness or the number of holes. These are not interchangeable senses of symmetry. Each belongs to a transformation group and a preserved structure. The word symmetry is therefore not praise for visual prettiness. It is a mathematical claim with a domain.

3. Similarity transforms: the operator beneath the entries

Under a change of basis, a matrix representation may become

$$A'=V^{-1}AV.$$

The entries of $A'$ can look nothing like those of $A$. Yet the characteristic polynomial survives:

$$\det(\lambda I-A') = \det(\lambda I-V^{-1}AV) = \det\left(V^{-1}(\lambda I-A)V\right).$$

Using multiplicativity of the determinant,

$$\det(\lambda I-A') = \det(V^{-1})\det(\lambda I-A)\det(V) = \det(\lambda I-A).$$

So the eigenvalues are preserved. The visible matrix changed. The linear operator, considered independently of coordinates, did not. This is one of the cleanest places to feel the distinction between an object and its representation.

4. Noether: when a symmetry becomes a conservation law

One of the deepest structural bridges in physics is Noether’s theorem: under the appropriate variational assumptions, a continuous symmetry is associated with a conserved quantity. The full machinery will be derived in Projection VIII. For now, take a simple case that can already be checked. Consider a free particle in one dimension with Lagrangian

$$L=\frac12m\dot q^2.$$

The coordinate $q$ does not appear in $L$. Translating the position,

$$q\rightarrow q+a,$$

changes nothing in the Lagrangian. The Euler–Lagrange equation has the form

$$\frac{d}{dt}\left(\frac{\partial L}{\partial\dot q}\right) - \frac{\partial L}{\partial q} =0.$$

Because

$$\frac{\partial L}{\partial q}=0,$$

we get

$$\frac{d}{dt}(m\dot q)=0.$$

Therefore

$$\boxed{p=m\dot q=\text{constant}.}$$

The translational symmetry has appeared as conservation of momentum. The deeper theorem is broader; the point here is that the bridge can be made operational. “Symmetry gives conservation” is not an occult slogan. In the right mathematical setting, the derivative actually vanishes. Time-translation symmetry is associated with energy conservation. Spatial translation with momentum. Rotation with angular momentum. And now the phrase earns its place:

$$\boxed{\textbf{SEEK WHAT SURVIVES TRANSFORMATION.}}$$

5. The glyph is allowed to be beautiful

A carved rosette, a tiling, a recursive triangle, a golden proportion, a celestial diagram—such forms can act as excellent glyphs. They compress relation into something the eye can carry. But a glyph is not a warrant. The golden ratio

$$\phi=\frac{1+\sqrt5}{2}$$

is genuine mathematics. Recursive geometry is genuine mathematics. Icosahedral symmetry is genuine mathematics. Their beauty may suggest a model or tell us where to look. It does not make them universal physical laws of optimization, consciousness, time, entropy, architecture, biology, or destiny. The question is always recoverable: What transformation is being claimed, and what has actually been shown to survive it? If that question cannot be answered, keep the beauty. Release the claim.

V — EVIDENCE

A scale receives a new weight. The balance moves.

A possibility space says what could happen. A probability model assigns weight to those possibilities. Evidence changes those weights. The world does not arrive whole. A finite observer receives traces: instrument readings, reports, images, memories, samples, signals, absences. The first discipline is Mentat-simple and difficult in practice: Do not confuse the observation with the inference you make from it.

0. Evidence enters through an observer

Partial measurement is only one source of incompleteness.

Before an observation becomes evidence, it must be selected, noticed, encoded, and interpreted.

Jeanette Andrews’ practice as an illusionist treats this sequence as live experimental material. Her work asks what happens when attention is directed, expectations are prepared, or a sensory event is made persuasive enough that the observer completes what was never fully given.

This does not mean that reality is arbitrary.

It means that the evidence channel has an architecture.

A useful audit therefore begins before Bayes:

ATTENTION — what entered awareness?
FRAMING — what possibilities were made salient before interpretation?
OBSERVATION — what was actually registered?
REPRESENTATION — how was it encoded?
INFERENCE — what conclusion was drawn from it?
TRUST — how much authority did the conclusion earn?

A magician can exploit this architecture locally. Contemporary media systems can exploit it at scale. Misinformation, synthetic media, and deepfakes make the same discipline urgent: a vivid percept is not yet a verified state of the world.

This is where Rayo’s concern with language and representation meets Andrews’ concern with attention and belief.

The observer receives a model-mediated view.

The view may be coherent.

The view may be compelling.

Neither property makes it complete.

Only now are we ready to update probabilities.

1. Bayes is a rotation of the joint probability

Let $H$ be a hypothesis and $E$ an observation treated as evidence. The joint probability can be factored as

$$P(H,E)=P(H)P(E\mid H).$$

The same joint probability can also be factored as

$$P(H,E)=P(E)P(H\mid E).$$

Set them equal:

$$P(H)P(E\mid H) = P(E)P(H\mid E).$$

Therefore

$$\boxed{ P(H\mid E) = \frac{P(E\mid H)P(H)}{P(E)}. }$$

Bayes’ theorem does not manufacture truth from a formula. It enforces consistency between two views of the same joint probability. The familiar words are useful if we remember that they name mathematical roles: $P(H)$ — prior probability. $P(E\mid H)$ — likelihood of the observed evidence under the hypothesis. $P(H\mid E)$ — posterior probability after conditioning on the evidence. $P(E)$ — the marginal probability that normalizes the possibilities. The object is the joint distribution. Bayes turns it.

2. Odds reveal what evidence actually does

For two hypotheses $H_1$ and $H_0$,

$$\boxed{ \frac{P(H_1\mid E)}{P(H_0\mid E)} = \frac{P(H_1)}{P(H_0)} \cdot \frac{P(E\mid H_1)}{P(E\mid H_0)}. }$$

The first factor is the prior odds. The second is the likelihood ratio—the degree to which the observation discriminates between the hypotheses. Suppose the prior odds are $1:9$ against $H_1$, but the observed evidence is twenty times more likely under $H_1$ than under $H_0$. Then

$$\frac{1}{9}\times20=\frac{20}{9}.$$

The posterior odds now favor $H_1$ by $20:9$. The evidence did not erase the prior. It multiplied it by its discriminatory force. This is disciplined revision rather than rhetorical confirmation. The question is not merely Does this observation fit my story? It is How much better did my story predict this observation than its rivals did?

3. Observation, measurement, model

The word evidence often compresses several different operations. An observation is what was seen, recorded, heard, counted, or directly reported. A measurement is an observation produced through an instrument or procedure with resolution, calibration, noise, and possible bias. A model links possible states of the world to expected observations. An inference moves from the observation back toward a claim about the state. These can fail separately. The observation can be misrecorded. The instrument can be noisy. The likelihood model can be wrong. The prior can be poorly chosen. The state space itself can omit the real explanation. A clean posterior cannot rescue a bad ontology.

4. Surprise becomes information

Using natural logarithms throughout the Tablet’s information-theory examples, information is measured in nats. If an event has probability $p$, one natural measure of its self-information is

$$I=-\ln p.$$

Rare events carry more surprise. For a discrete random variable $X$ with distribution $p(x)$, the expected self-information is the Shannon entropy:

$$\boxed{ H(X) = -\sum_x p(x)\ln p(x). }$$

This is information theory. It is not automatically thermodynamic entropy. The logarithm will return near the end of the Tablet, where we will show exactly when the mathematical bridge to Boltzmann counting can be made—and exactly what that bridge does not permit us to say.

5. Measurement and model negotiate through uncertainty

Suppose a scalar state has prior estimate $\hat x^{-}$ with prior variance $\Sigma^{-}$. We observe

$$z=x+v,$$

where the measurement noise $v$ has variance $R$. For this simple direct-observation case, the Kalman gain is

$$K = \frac{\Sigma^{-}}{\Sigma^{-}+R}.$$

The updated estimate is

$$\boxed{ \hat x^{+} = \hat x^{-}+K(z-\hat x^{-}). }$$

The difference

$$z-\hat x^{-}$$

is the innovation: what arrived that the prediction did not already contain. The updated variance is

$$\boxed{ \Sigma^{+}=(1-K)\Sigma^{-}. }$$

If the prior is uncertain relative to the measurement, $K$ moves toward $1$: the estimate follows the measurement strongly. If the measurement is noisy relative to the prior, $K$ becomes small: the estimate moves cautiously. The update is not “trust the data” versus “trust the model.” It is a weighted negotiation whose weights depend on quantified uncertainty—under the assumptions of the model. That last clause matters. A filter can become exquisitely confident inside the wrong model.

6. Uncertainty is not one substance

Probability may represent uncertainty about which state is true. Measurement uncertainty may represent instrument limitations. Parameter uncertainty may represent imperfect knowledge of model values. Model uncertainty may mean that the structure itself is wrong or incomplete. Quantum theory introduces a different category again. For position and momentum,

$$\Delta x\,\Delta p \ge \frac{\hbar}{2}.$$

That relation belongs to quantum mechanics and has a specific mathematical and physical meaning. It is not a poetic license for ordinary indecision, social ambiguity, or poorly specified data. A disciplined reader does not gather everything called “uncertainty” into one mystical bucket. The reader asks: What mathematical structure produces this uncertainty, and what would reduce it—if anything can?

7. Evidence changes knowledge, not jurisdiction

A better model can improve prediction. A better sensor can improve estimation. A stronger likelihood ratio can rationally change confidence. None of those operations grants permission. A person can know more and still lack authority to act. A system can predict consent and still lack consent. A model can rank an intervention and still have no jurisdiction to impose it. Knowledge changes the epistemic state. It does not automatically change who may decide. That boundary will matter more once inference enters a feedback loop. Before action, however, another question has to be learned: When something changes here, where did it come from—and where did it go?

VI — CONSERVATION & FLOW

A vessel, a boundary, a current.

A conservation law is not merely an equation saying that something stays the same. It is an account of what changes inside a boundary and what crosses it. The boundary is not a decorative line around the mathematics. It decides what the ledger can see.

1. Begin with a region

Let $\rho(\mathbf x,t)$ be the density of some quantity inside a fixed spatial region $V$. The amount stored in the region is

$$Q_V(t)=\int_V \rho(\mathbf x,t)\,dV.$$

Let $\mathbf J$ be the flux of that quantity, $\mathbf n$ the outward unit normal on the boundary $\partial V$, and $s$ a local source density. The rate of change of what is stored equals what is produced inside minus what flows outward:

$$\boxed{ \frac{d}{dt}\int_V \rho\,dV = -\int_{\partial V}\mathbf J\cdot\mathbf n\,dA +\int_V s\,dV. }$$

Use the divergence theorem:

$$\int_{\partial V}\mathbf J\cdot\mathbf n\,dA = \int_V \nabla\cdot\mathbf J\,dV.$$

Then

$$\int_V \left( \frac{\partial\rho}{\partial t} +\nabla\cdot\mathbf J -s \right)dV =0.$$

If this holds for arbitrary regions, the integrand must vanish:

$$\boxed{ \frac{\partial\rho}{\partial t} +\nabla\cdot\mathbf J =s. }$$

The global ledger has become a local law. The integral and differential forms are not rival descriptions. They are the same balance seen at two scales.

2. The nouns can change while the grammar survives

For mass, $\rho$ may be mass density. For electric charge, charge density. For probability, a probability density over position or another state space. For traffic, vehicle density along a roadway. The same grammar can recur: storage — flux — source — sink — boundary. But a repeated grammar does not merge the nouns. Cars are not molecules. Probability is not mass. Charge is not traffic. A structural correspondence can be exact while the physical meanings remain entirely different. That distinction is one of the Tablet’s deepest protections. Transfer the mathematics when the mapping warrants it. Do not smuggle ontology through the same doorway.

3. The boundary decides what looks conserved

Move the boundary and the bookkeeping changes. A quantity that appears to vanish from a small region may simply have crossed its edge. A local subsystem can become more ordered while exporting heat or entropy elsewhere. A financial model can improve one account by moving cost to another. A population average can rise while a subgroup absorbs the loss. Those last examples are not claims that social cost is a thermodynamic flux. They share a narrower structural lesson: a boundary can hide a receiver. The scientific form of the lesson is precise: in a conservation law, flux through $\partial V$ matters.

The wider discipline is equally precise: if a ledger improves after the boundary is redrawn, inspect what was moved outside it.

4. Symmetry returns as bookkeeping

In Projection IV, conservation appeared through symmetry. Here it appears through local accounting. Those are different approaches to related structure. Noether asks: What continuous symmetry implies a conserved quantity? The continuity equation asks: How does that quantity move through space once a local density and flux are defined? One route begins with invariance. The other begins with a boundary. They meet in conservation without becoming the same theorem.

5. Entropy is a balance with production

Thermodynamic entropy requires special care because it is not generally conserved. A local entropy balance may be written schematically as

$$\boxed{ \frac{\partial s_{\!\mathrm{th}}}{\partial t} +\nabla\cdot\mathbf J_s = \sigma_s, \qquad \sigma_s\ge0, }$$

where $s_{\!\mathrm{th}}$ is an entropy density, $\mathbf J_s$ an entropy flux, and $\sigma_s$ local entropy production. The production term changes the story. Entropy is not simply a conserved fluid that sloshes from one box to another. Irreversible processes can produce it. A local region can become more ordered while energy is consumed and entropy is exported to the surroundings. Geometry can redirect heat, matter, current, stress, radiation, vibration, or information flow. It does not abolish thermodynamic entropy because the shape is clever. And the word entropy does not become a scientific unit for confusion, decadence, disorderly management, or a bad Tuesday merely because the metaphor is vivid.

The equation has a domain. Its meaning lives there.

6. The scribe’s ledger

Conservation teaches a habit older than modern notation: if something increased, ask what entered or what was produced; if something decreased, ask what left or what was destroyed; if a local region became orderly, inspect the boundary; if the total stayed fixed, inspect the redistribution; if a cost vanished after the frame changed, ask where the frame put it. The ledger does not accuse. It refuses disappearance by syntax. That habit will return when the Tablet changes scale. First, we have to let local relations travel.

VII — FIELDS, WAVES & TRANSPORT

A local disturbance learns how to cross space.

An isolated oscillator can move forever without sending anything anywhere. A wave requires coupling. One location must matter to another. The local becomes extended through a rule.

1. The wave equation

A canonical scalar wave equation is

$$\boxed{ \frac{\partial^2u}{\partial t^2} = c^2\nabla^2u. }$$

The left side measures local acceleration in time. The right side responds to spatial curvature. A disturbance creates gradients. The gradients influence neighboring points. The coupling allows a pattern to propagate. A global wave emerges from local relations.

2. Euler returns as a traveling mode

In one dimension, try the complex mode

$$u(x,t)=e^{\mathrm{i}(kx-\omega t)}.$$

Differentiate in time:

$$\frac{\partial^2u}{\partial t^2} =-\omega^2u.$$

Differentiate in space:

$$\frac{\partial^2u}{\partial x^2} =-k^2u.$$

Insert both into

$$u_{tt}=c^2u_{xx}.$$

Then

$$-\omega^2u=-c^2k^2u,$$

so a nonzero mode requires

$$\boxed{\omega^2=c^2k^2.}$$

The complex exponential from Projection III has acquired a spatial phase. Take the real part:

$$\Re\,u(x,t)=\cos(kx-\omega t).$$

The rotation is traveling. What was once a point moving around a circle has become a phase moving through space. The mathematics performs the turn. The physical meaning comes from the wave equation that admitted the mode.

3. Fourier becomes computational machinery

Because the wave equation is linear, sums of solutions are again solutions. That means a complicated initial shape can be decomposed into Fourier modes, each mode can be evolved according to the equation, and the result can be recombined. In physical space, differentiation couples nearby values and may make the evolution look complicated. In Fourier space, spatial derivatives become multiplication by powers of $ik$. For a Fourier mode,

$$\frac{\partial}{\partial x}e^{\mathrm{i}kx}=\mathrm{i}k e^{\mathrm{i}kx},$$

and

$$\frac{\partial^2}{\partial x^2}e^{\mathrm{i}kx}=-k^2e^{\mathrm{i}kx}.$$

The operator has become algebraic on the mode. This is why

$$\boxed{\textbf{SAME OBJECT, DIFFERENT BASIS.}}$$

is not decorative philosophy. It is a method for making difficult operations simple.

4. Same basis, different generator, different physics

Now replace the wave equation by the diffusion equation:

$$\boxed{ \frac{\partial u}{\partial t} = D\nabla^2u. }$$

Try a spatial Fourier mode with time-dependent amplitude:

$$u(\mathbf x,t)=a_{\mathbf k}(t)e^{\mathrm{i}\mathbf k\cdot\mathbf x}.$$

Because

$$\nabla^2e^{\mathrm{i}\mathbf k\cdot\mathbf x} =-|\mathbf k|^2e^{\mathrm{i}\mathbf k\cdot\mathbf x},$$

the diffusion equation gives

$$\dot a_{\mathbf k} =-D|\mathbf k|^2a_{\mathbf k}.$$

Therefore

$$\boxed{ a_{\mathbf k}(t) = a_{\mathbf k}(0)e^{-D|\mathbf k|^2t}. }$$

High-frequency modes decay faster because of the $|\mathbf k|^2$ factor. Compare that with the wave equation, whose modes oscillate rather than simply decay. The spatial basis can be identical. The generator of time evolution is not. Therefore

$$\boxed{ \text{same representation} +\text{different generator} =\text{different dynamics}. }$$

And, when the models describe physical systems, different dynamics mean different physics. This is exactly why mathematical similarity must not be confused with physical identity.

5. Black–Scholes rotates into diffusion

A strong correspondence should survive being performed in public. For the standard Black–Scholes model of a non-dividend-paying asset, the option value $V(S,t)$ satisfies

$$\frac{\partial V}{\partial t} +\frac12\sigma^2S^2\frac{\partial^2V}{\partial S^2} +rS\frac{\partial V}{\partial S} -rV =0.$$

Introduce time-to-maturity

$$\tau=T-t$$

and logarithmic price

$$x=\ln\left(\frac{S}{K}\right), \qquad V(S,t)=K\,v(x,\tau).$$

Then

$$S\frac{\partial V}{\partial S}=K v_x$$

and

$$S^2\frac{\partial^2V}{\partial S^2}=K(v_{xx}-v_x).$$

Also,

$$\frac{\partial V}{\partial t}=-K v_\tau.$$

Substitute these into the pricing equation and divide by $K$:

$$-v_\tau +\frac12\sigma^2(v_{xx}-v_x) +rv_x-rv =0.$$

So

$$\boxed{ v_\tau = \frac12\sigma^2v_{xx} +\left(r-\frac12\sigma^2\right)v_x -rv. }$$

Define

$$\mu=r-\frac12\sigma^2$$

and set

$$v(x,\tau)=e^{ax+b\tau}u(x,\tau).$$

Then

$$v_x=e^{ax+b\tau}(au+u_x),$$

$$v_{xx}=e^{ax+b\tau}(a^2u+2au_x+u_{xx}),$$

and

$$v_\tau=e^{ax+b\tau}(bu+u_\tau).$$

Choose

$$a=-\frac{\mu}{\sigma^2}$$

to cancel the first-derivative term, and choose

$$b=-r-\frac{\mu^2}{2\sigma^2}$$

to cancel the remaining zeroth-order term. What remains is

$$\boxed{ u_\tau=\frac12\sigma^2u_{xx}. }$$

Rescale time once more:

$$\eta=\frac12\sigma^2\tau.$$

Then

$$\boxed{u_\eta=u_{xx}.}$$

The Black–Scholes equation has been transformed into the heat equation. Now the boundary: prices did not become temperatures; markets did not become molecules; financial risk did not become thermal energy. What transferred was mathematical structure. Once the mapping is established, methods developed for one differential equation family can help solve another. This is correspondence at full strength: strong enough to move technique; not strong enough to move ontology.

6. Nonlinearity breaks the easy modal story

For an incompressible Newtonian fluid, a common form of the Navier–Stokes equations is

$$\boxed{ \frac{\partial\mathbf u}{\partial t} +(\mathbf u\cdot\nabla)\mathbf u = -\frac{1}{\rho}\nabla p +\nu\nabla^2\mathbf u +\mathbf f, }$$

with

$$\nabla\cdot\mathbf u=0.$$

Here $\nu$ is kinematic viscosity. The nonlinear advection term

$$(\mathbf u\cdot\nabla)\mathbf u$$

changes everything. In a linear system, Fourier modes can often evolve independently. With nonlinearity, modes interact. Energy can move among scales. A basis that diagonalizes the linear part does not generally diagonalize the full dynamics. Representation can reveal structure. It cannot abolish structure that is really there. That is another kind of intellectual honesty: a powerful method should know when its own simplification stops working. The field has shown us propagation, diffusion, transformed equation families, and nonlinear coupling. Now we ask a different question. Among all admissible paths, how can a law be written as a condition on the path itself?

VIII — CONSTRAINT & EXTREMUM

A path appears because all other nearby paths were allowed to argue with it.

Possibility tells us which states are admissible. Dynamics tells us how a state changes locally. A variational principle rotates the description again: instead of following the motion instant by instant, compare whole candidate paths.

1. The action is a function of a function

Let a path be $q(t)$ and define the action

$$\boxed{ S[q] = \int_{t_0}^{t_1}L(q,\dot q,t)\,dt. }$$

$S$ is a functional: it takes an entire function $q(t)$ as its input and returns a number. Perturb the path:

$$q(t)\rightarrow q(t)+\varepsilon\eta(t),$$

with fixed endpoints,

$$\eta(t_0)=\eta(t_1)=0.$$

For a stationary path,

$$\left. \frac{d}{d\varepsilon}S[q+\varepsilon\eta] \right|_{\varepsilon=0} =0.$$

Differentiate under the integral:

$$\int_{t_0}^{t_1} \left( \frac{\partial L}{\partial q}\eta + \frac{\partial L}{\partial\dot q}\dot\eta \right)dt =0.$$

Integrate the second term by parts:

$$\int_{t_0}^{t_1} \frac{\partial L}{\partial\dot q}\dot\eta\,dt = \left[ \frac{\partial L}{\partial\dot q}\eta \right]_{t_0}^{t_1} - \int_{t_0}^{t_1} \frac{d}{dt} \left(\frac{\partial L}{\partial\dot q}\right) \eta\,dt.$$

The boundary term vanishes because $\eta$ vanishes at both endpoints. Therefore

$$\int_{t_0}^{t_1} \left[ \frac{\partial L}{\partial q} - \frac{d}{dt} \left(\frac{\partial L}{\partial\dot q}\right) \right]\eta(t)\,dt =0.$$

Because the variation $\eta(t)$ is otherwise arbitrary,

$$\boxed{ \frac{d}{dt} \left(\frac{\partial L}{\partial\dot q}\right) - \frac{\partial L}{\partial q} =0. }$$

The Euler–Lagrange equation has emerged. A global statement about stationary action has become a local differential equation. This is not merely a claim that two descriptions are connected. We just performed the rotation.

2. The oscillator returns through a different door

Take the harmonic oscillator with

$$L(q,\dot q) = \frac12m\dot q^2- \frac12kq^2.$$

Then

$$\frac{\partial L}{\partial\dot q}=m\dot q$$

and

$$\frac{\partial L}{\partial q}=-kq.$$

Euler–Lagrange gives

$$\frac{d}{dt}(m\dot q)-(-kq)=0,$$

so

$$\boxed{m\ddot q+kq=0.}$$

Divide by $m$ and define

$$\omega^2=\frac{k}{m}.$$

Then

$$\ddot q+\omega^2q=0,$$

which is the same oscillator equation whose imaginary exponential modes appeared in Projection III. There, oscillation arrived through complex roots and rotation. Here, it arrived through stationary action. Same dynamics. Different basis of thought.

3. Noether’s window opens fully

Now the earlier symmetry result has somewhere to stand. If a coordinate $q$ does not appear explicitly in $L$, then

$$\frac{\partial L}{\partial q}=0.$$

Euler–Lagrange becomes

$$\frac{d}{dt} \left( \frac{\partial L}{\partial\dot q} \right)=0.$$

Therefore the conjugate momentum

$$p_q=\frac{\partial L}{\partial\dot q}$$

is conserved:

$$\boxed{\dot p_q=0.}$$

The symmetry is not decorative. The variational structure converts it into a conserved quantity. This is the kind of relation the Tablet is built to recognize: not because two words rhyme, but because a transformation changes one mathematical description into another while preserving the underlying dynamics.

4. Global does not mean intentional

The variational form can make nature sound teleological, as if a system surveys all possible futures and selects the most elegant one. The mathematics requires no such story. Under the appropriate assumptions, the global stationary-action formulation and the local differential equations encode the same dynamics. A compact global criterion is not evidence of foresight. Formal economy is not agency. A beautiful formulation can be exact without being a mind.

5. Constraint changes the space of admissible directions

Suppose we want to extremize $f(\mathbf x)$ subject to

$$g(\mathbf x)=c.$$

At a regular point where $\nabla g\neq0$, the allowed infinitesimal directions lie tangent to the constraint surface, while $\nabla g$ is normal to it. At a constrained extremum, $\nabla f$ can have no component along an allowed tangent direction. Therefore it must be parallel to $\nabla g$:

$$\boxed{ \nabla f=\lambda_c\nabla g. }$$

The Lagrange multiplier $\lambda_c$ is not a magical extra variable. It records the fact that optimization is happening in a restricted geometry. Constraint changes what counts as a possible direction of improvement. That is why an unconstrained optimum can be meaningless in the real problem.

6. Every optimum hides a value judgment somewhere upstream

Optimized for what? Over which variables? Under what constraints? Over what horizon? At whose cost? Under what uncertainty? A minimum is always a minimum of something. A maximum is always a maximum over some feasible set. Mathematics can expose an optimum once the objective and constraints are specified. It does not choose the objective’s moral authority. The equation can answer what follows if these are the rules? It cannot answer who had the right to choose the rules? That boundary will become impossible to ignore when action enters the dynamics.

IX — ACTION & FEEDBACK

The hand touches the mechanism and becomes one of its causes.

A model predicts. An agent acts. The next observation comes from a world partly changed by that action. The loop closes.

1. Put an input into the state equation

Return to

$$\dot{\mathbf x}=A\mathbf x.$$

Now permit intervention:

$$\boxed{ \dot{\mathbf x} =A\mathbf x+B\mathbf u. }$$

Let the measured output be

$$\boxed{ \mathbf y=C\mathbf x+D\mathbf u. }$$

The state is not necessarily the observation. A sensor may see only part of a system. A dashboard can display only selected coordinates. An estimate can be wrong while every visible digit is precise. The measurement channel is a projection. That fact has been with us since the first page.

2. Feedback changes the generator

Suppose full-state feedback is chosen as

$$\mathbf u=-K\mathbf x.$$

Then

$$\dot{\mathbf x} =A\mathbf x-BK\mathbf x,$$

so

$$\boxed{ \dot{\mathbf x}=(A-BK)\mathbf x. }$$

Therefore

$$\boxed{ \mathbf x(t) =e^{(A-BK)t}\mathbf x(0). }$$

The matrix exponential has returned, but not in the same role. Earlier, $A$ generated the evolution; feedback has now changed the generator to

$$A-BK.$$

The intervention is not external commentary on the system. It has entered the dynamics.

3. In one dimension the change is impossible to hide

Take the scalar system

$$\dot x=ax+bu$$

with feedback

$$u=-kx.$$

Then

$$\dot x=(a-bk)x,$$

so

$$\boxed{ x(t)=e^{(a-bk)t}x(0). }$$

The sign of $a-bk$ determines whether this linear closed-loop mode grows or decays. A controller changes stability by changing the effective generator. That is the structure beneath the larger matrix equation.

4. The eigenvalues return as stability

For the multivariable linear system, the eigenvalues of $A-BK$ determine the modal behavior of the closed loop. The same eigenvalue idea that first appeared as a way to decouple dynamics now appears as a way to judge whether perturbations decay, persist, or grow. Representation and evolution have both returned, but action has changed the operator they belong to. This is how the Tablet should recur: not by repeating a slogan, but by forcing an old structure to do new work.

5. Estimation enters the loop

If the full state cannot be observed directly, the controller may act on an estimate $\hat{\mathbf x}$ instead. The earlier architecture returns:

$$\text{MODEL} + \text{MEASUREMENT} + \text{UNCERTAINTY} \longrightarrow \text{REVISED ESTIMATE}.$$

The estimate informs action; action changes the system; and the changed system produces the next measurement. The larger loop is

$$\boxed{ \text{STATE} \rightarrow \text{CHANGE} \rightarrow \text{OBSERVATION} \rightarrow \text{INFERENCE} \rightarrow \text{ACTION} \rightarrow \text{CONSEQUENCE} \rightarrow \text{NEW STATE}. }$$

This is an architectural loop, not a universal differential equation. It becomes literal only when those pieces are actually defined in a particular system.

6. A controller can fail by succeeding

Suppose the model omits a variable that matters. Suppose an actuator saturates. Suppose there is a delay. Suppose the gain works in the operating regime used for design but destabilizes another regime. Suppose the measured output is merely a proxy for the real objective. Then the controller can optimize the model while damaging the world the model was meant to represent. This is one of the cruelest failures in quantitative reasoning because the equations may remain internally immaculate. The model can be right about itself. The representation can still be wrong about the system.

7. Capability is not permission

A system may be controllable. An operator may know exactly how to control it. Neither statement grants authority.

Capability is not permission.

Knowledge does not create authority.

The mathematics can show that an intervention is possible. It cannot manufacture jurisdiction from possibility. And the assumption of a single controller becomes even less adequate when the environment contains other agents who choose back.

X — OTHER AGENTS

Another hand reaches into the mechanism—and does not belong to you.

So far, action has been privileged. The control input $\mathbf u$ belonged to one controller. But many environments contain other centers of choice. Then one agent’s action is another agent’s condition. The system becomes strategic.

1. Why the single-controller picture breaks

Start with a dynamical system influenced by two decision-makers:

$$\boxed{ \dot{\mathbf x} =A\mathbf x+B_1\mathbf u_1+B_2\mathbf u_2. }$$

If agent 1 chooses $\mathbf u_1$, it still does not determine the trajectory by itself because agent 2 chooses $\mathbf u_2$. The second input is not necessarily noise. It can be purposeful. It can depend on the same state:

$$\mathbf u_2=\pi_2(\mathbf x,\text{information available to agent 2}).$$

Now agent 1’s best action can depend on what agent 2 is expected to do, while agent 2’s best action can depend on agent 1. The control problem has changed form. There is no longer one objective function with one privileged hand on the lever.

2. From control to game

Let player $i$ choose strategy $s_i$, and let $s_{-i}$ denote the strategies of the others. Player $i$ receives payoff

$$u_i(s_i,s_{-i}).$$

A strategy profile $s^*$ is a Nash equilibrium if no player can improve by changing strategy unilaterally:

$$\boxed{ u_i(s_i^*,s_{-i}^*) \ge u_i(s_i,s_{-i}^*) }$$

for every available unilateral deviation $s_i$. Equivalently,

$$s_i^*\in\operatorname*{arg\,max}_{s_i}u_i(s_i,s_{-i}^*)$$

for each player $i$ when the maximization is well defined. The phrase doing the work is: given what the others are doing. Equilibrium is a mutual consistency condition among best responses. It is not a halo.

3. Equilibrium is not goodness

Consider a Prisoner’s-Dilemma-style payoff table:

• cooperate / cooperate: $(3,3)$
• defect / cooperate: $(5,0)$
• cooperate / defect: $(0,5)$
• defect / defect: $(1,1)$

If defection is each player’s best response regardless of the other’s choice, then

$$(\text{defect},\text{defect})$$

is a Nash equilibrium. Yet both players receive more at

$$(\text{cooperate},\text{cooperate}).$$

So equilibrium does not mean: fair; collectively optimal; consensual; morally legitimate; or robust to a change in the rules. It means exactly what the equilibrium condition says. No more.

4. Mechanism design rotates the question again

Game theory can ask: Given these rules, what strategies are stable? Mechanism design asks: What rules change the incentives, information, or feasible actions? A disclosure rule changes information. An interlock removes an action. A payment rule changes payoffs. A refusal right changes the feasible set. A protocol changes what counts as a valid move. Change the mechanism and the game changes. Constraint, information, action, and other agents have now entered the same object.

5. Other agents are not disturbances

Engineering language can be dangerous when it escapes its domain. An unmodeled physical input may be called a disturbance. Another person is not merely noise in one controller’s objective function. They have their own information, preferences, constraints, uncertainty, and authority. A prediction that someone will agree is not agreement. A high probability of consent is not consent. A mechanism that makes refusal expensive may be stable and still coercive. A system can be strategically stable and ethically unacceptable. The Tablet must therefore preserve three separations: knowledge; capability; authority. Mathematics can illuminate the first two. It does not automatically supply the third.

6. Compression can erase the receiver

A scalar objective can improve while one participant absorbs the cost. A global average can rise while a subgroup worsens. A model can describe everyone’s actions while omitting someone’s right to refuse. The optimization algorithm may be flawless. The state description may be the failure. The payoff function may be the failure. The feasible set may be the failure. The boundary may be the failure. So a new question enters the Tablet: Who disappeared when the system was compressed? That question is larger than game theory. It is a question about scale.

XI — SCALE, ENTROPY & RETURN

The same mark seen again after the journey, carrying wear.

A system seen from far away is not merely a smaller picture of the same description. Scale changes what counts as state. Detail disappears. Regularities emerge. And some transformations cannot be undone from the information that remains.

1. Coarse-graining is a many-to-one map

Let a microscopic state be $\xi$. Let the macroscopic description be

$$M=\Pi_{\mathrm{cg}}(\xi).$$

Many different microstates may map to the same macrostate:

$$\xi_1,\xi_2,\ldots,\xi_N \longmapsto M.$$

The map $\Pi_{\mathrm{cg}}$ is therefore generally many-to-one. Once only $M$ is retained, the exact identity of $\xi$ may no longer be recoverable. That is information loss produced by representation. The transformation asks the Tablet’s oldest question again: What survived coarse-graining?

2. Counting returns as thermodynamic multiplicity

In the microcanonical setting, suppose a macrostate is compatible with $\Omega$ microscopic configurations. Boltzmann’s relation is

$$\boxed{ S=k_B\ln\Omega. }$$

Why the logarithm? Because independent multiplicities multiply. If two independent subsystems have multiplicities $\Omega_1$ and $\Omega_2$, then

$$\Omega_{12}=\Omega_1\Omega_2.$$

Apply the logarithm:

$$\begin{aligned} S_{12} &=k_B\ln(\Omega_1\Omega_2)\\ &=k_B\ln\Omega_1+k_B\ln\Omega_2\\ &=S_1+S_2. \end{aligned}$$

Thus

$$\boxed{\text{multiplicative counting}\longrightarrow\text{additive entropy}.}$$

The logarithm is not decorative. It performs the conversion. Counting has returned from Projection I carrying an entirely new physical interpretation.

3. Shannon meets Boltzmann—under stated assumptions

Recall Shannon entropy:

$$H(X) = -\sum_x p(x)\ln p(x).$$

Now consider an idealized uniform distribution over $\Omega$ accessible microstates:

$$p_i=\frac{1}{\Omega} \qquad (i=1,\ldots,\Omega).$$

Then

$$\begin{aligned} H &=-\sum_{i=1}^{\Omega}\frac1\Omega \ln\left(\frac1\Omega\right)\\ &=-\Omega\frac1\Omega(-\ln\Omega)\\ &=\ln\Omega. \end{aligned}$$