the second hundred · metaphor 164

Up close, only
the local slope.

Nudge someone a little — a small ask, a small slight — and what happens next has almost nothing to do with their whole life story. It depends on the slope of their character right where you touched it: whether a small push here grows into something, or quietly fades.

We keep trying to read people globally — their history, their values, the arc of who they've been. But behaviour in the moment is a local thing. Two people with the same grand story can respond to the identical nudge in opposite ways, because one was standing on a floor that pushes back and the other on a ridge that lets go. The story is the same; the ground underfoot is not.

Mathematics has an exact object for this: the Jacobian, the best straight-line approximation to a curved response, right at the point you're standing. It stretches, rotates, and flips small perturbations, and two numbers hidden inside it — its eigenvalues — decide the fate of every small nudge: grow, or fade.

the flow of a character · drag the probe x = mood · y = momentum
the probe · where you nudge Jacobian ellipse · the linear image true flow · the real nonlinear image eigen-directions
the Jacobian J at the probe
0.000.00 0.000.00
δnudge′ = J · δnudge
trace (tr)
det
eigenvalue λ₁
eigenvalue λ₂
Drag the probe over the field.
Temperament · damping δ (how much the character bleeds off momentum)0.30
−0.6 · self-amplifying0 · conservative+1.4 · steadying
Nudge size · radius of the perturbation ring0.35
Drag the probe onto a marked point, or anywhere between.
Presets
Every point in the field has its own Jacobian. Drag the probe onto a settled self at (±1, 0), then onto the ridge at (0, 0) — same character, opposite fate for a nudge.

The idea

Zoom in far enough and every curve is a line.

A person's response to the world is a curved, tangled thing — nonlinear, in the jargon. But zoom in on any one point and the tangle straightens out. In an infinitesimal neighbourhood, the entire response collapses to a single linear map: the Jacobian. It is a matrix, and a matrix does only three things to a small arrow — stretch it, rotate it, flip it. That's the whole local repertoire. A tiny circle of possible nudges around the point comes back, a moment later, as an ellipse: some directions blown up, some squashed, maybe reflected.

Which nudges grow and which fade is not written on the ellipse's face but in the matrix's eigenvalues — the special directions the map only scales, without turning. A positive eigenvalue means a nudge along that direction is amplified; a negative one means it decays; a complex pair means it spirals. The signs alone classify the point: all-fading is a stable self, all-growing is unstable, and one of each is a saddle — stable along one axis, precarious along another.

None of this needs the global story. The Jacobian is built from derivatives evaluated at one point; it knows nothing of where the person has been or where the flow eventually carries them. It answers exactly one question, and answers it exactly: if I push a little, right here, does it grow or fade?

What to try

Drag the probe. Watch the circle become an ellipse.

The field is a real dynamical system — a mood x with a momentum y, curving under its own rule. Wherever you drop the probe, the instrument computes the Jacobian from the actual derivatives at that point and draws its action: the faint ring is a spray of small nudges; the bright ellipse is where the linear map sends them; the dotted true-flow dots are where the real, curved system sends them. For a small ring the two agree almost perfectly — that agreement is the theorem. Grow the nudge size and watch them peel apart: the linear picture is only local.

Land on a settled self at (1, 0): the ellipse shrinks inward, eigenvalues negative, a nudge fades. Land on the ridge at (0, 0): the ellipse stretches along one axis and shrinks along the other — a saddle, one eigenvalue each sign. Now drive the temperament slider: at δ = 0 the settled selves become pure centers, eigenvalues on the imaginary axis, nudges neither grow nor fade but circle forever. Push δ negative and those same points flip to self-amplifying spirals. The story never changed. Only the slope did.

The mapping

The local truth of a person.

We are quick to explain a reaction by the biography: of course he snapped, look at his childhood. The Jacobian is the discipline of the opposite move. It says the reaction to this nudge, now, is governed by the local structure — the slope of the response surface at the exact point of contact — and that structure can differ wildly from one point to the next along the same life. The calm colleague who absorbs a hard note today (eigenvalues negative, disturbance fades) can be standing, next week, on a saddle where the identical note runs away from him.

It also names why identical people diverge under identical pushes: not different souls, but different eigen-directions. Push a saddle along its stable direction and nothing happens; push it a hair off, along the unstable one, and it bolts. The content of the nudge matters less than its alignment with the local geometry. Which is why the same sentence lands as nothing from one angle and as a rupture from another — and why paying attention to where someone is, not just who they are, is the whole art.

Read as life lessons

Three things the local slope knows.

01

Behaviour is local

The response to a small nudge lives in a derivative, not a résumé. Same person, different point on the curve, opposite reaction — and both are true.

02

Direction beats size

A saddle is safe along one axis and fatal along another. What decides the outcome is how a push aligns with the eigen-directions — not how hard you pushed.

03

The sign is the fate

You don't need to solve the whole trajectory. The eigenvalue signs at the point already say grow, fade, or spiral. The linearization decides the small.

The mapping, exactly

Mathematics ↔ life.

MathematicsLife
the nonlinear flowA whole character — the tangled, curved way a person actually responds across every situation.
the point x*Exactly where they are right now: this mood, this moment, the spot you happen to be touching.
the Jacobian JThe local slope of their character — the best linear read of how they meet a small push, here.
a perturbation δA small nudge — an ask, a slight, a kindness — small enough that only the local slope governs it.
eigenvalues Re(λ)Whether that nudge grows or fades, and how fast — the fate written into the point itself.
eigen-directionsThe axes along which they're brittle or unshakable; the same push lands differently depending on alignment.
saddle pointSteady in one respect, one nudge from bolting in another — the person poised between two selves.

The honest model

What's really under the hood.

The field is an unforced Duffing oscillator: ẋ = y, ẏ = x − x³ − δy. It has three fixed points — two settled selves at (±1, 0) and a ridge (saddle) at (0, 0). Its Jacobian is J = [[0, 1], [1 − 3x², −δ]], and the page evaluates those entries directly at wherever you drop the probe. Nothing is tabulated: the trace, determinant, eigenvalues λ = (tr ± √(tr²−4·det))/2, and the classification are all recomputed from that live matrix.

The bright ellipse is the honest image of a small ring under the linear map exp(Jτ) — computed in closed form for the 2×2, no lookup — while the dotted cloud is the same ring pushed through the full nonlinear equations by a Runge–Kutta integrator. Watching them coincide for a small ring and separate for a large one is the linearization theorem happening in front of you, not asserted. The eigen-direction lines are the real eigenvectors (1, λ) of J, drawn only when the eigenvalues are real.

Where the metaphor tears

Three honest failures.

It is only true for small nudges.

The Jacobian is a first-order truth: exact in the limit, approximate for anything you'd actually notice. Grow the ring in the instrument and the linear ellipse drifts off the real flow. A big enough shove leaves the neighbourhood where the local slope rules, and the global shape of the person — the very thing the Jacobian ignores — takes over. Do not read a large blow off a linear map.

Zero eigenvalue, no verdict.

When an eigenvalue sits exactly on the line between growing and fading — a center, a fold — the linearization goes mute. The nudge neither clearly grows nor clearly fades, and what actually happens is decided by the higher-order terms the Jacobian threw away. The most interesting human moments — the ones that could tip either way — are precisely where this crude first look tells you least.

A person is not a fixed rule.

The Duffing field never changes; a person's does, every day, from sleep and hunger and who last spoke to them. The Jacobian assumes the map holds still long enough to differentiate it. Sometimes it doesn't — the slope you carefully measured this morning is a different slope by afternoon, and the elegant classification you drew is already stale.