A number that does two jobs
By now f'(z_0) means something precise to you: the limit of the difference quotient (f(z) - f(z_0)) / (z - z_0), a single complex number that every direction of approach must agree on. That is the algebra. This guide is about the picture behind that number — what f'(z_0) actually does to the plane near z_0. The short answer, and the one to carry away, is that it rotates and scales every tiny arrow leaving z_0, all at once, by the same amount.
The clue is hiding in how complex multiplication works, which you met long ago. Write any complex number in polar form: f'(z_0) = r e^(i theta), where r = |f'(z_0)| is its modulus and theta = arg f'(z_0) is its argument. Multiplying any other complex number by r e^(i theta) does exactly two things to it — stretches its length by the factor r, and rotates it by the angle theta. One complex number, one stretch, one turn. Keep that fact warm; the whole guide is just this fact applied to the derivative.
Now feed this into the local picture of complex differentiability. Near z_0, a holomorphic function is well approximated by f(z) ≈ f(z_0) + f'(z_0)(z - z_0). The interesting part is the displacement z - z_0 — the little arrow pointing from z_0 to a nearby z. Multiplying that arrow by f'(z_0) is, by the polar fact, exactly a stretch by |f'(z_0)| and a turn by arg f'(z_0). So to first order the map takes the arrow from z_0 to z and turns it into a stretched, rotated arrow from f(z_0) to f(z). That is the complex derivative doing its geometric work.
Amplitwist: amplify and twist
Tristan Needham gave this combined action a name that sticks once you hear it: the amplitwist. Fuse 'amplify' (stretch the length) with 'twist' (rotate the direction) and you have the one verb that describes what f'(z_0) does. Its modulus |f'(z_0)| is the amplification — how much longer every small arrow gets — and its argument arg f'(z_0) is the twist — the angle every small arrow turns through. The two numbers a single complex derivative carries are precisely a stretch factor and a rotation angle, bundled together.
Concretely: take any tiny arrow leaving z_0, of length L and pointing at angle phi. Multiplying it by f'(z_0) = r e^(i theta) sends it to an arrow of length r times L (the amplify step) pointing at angle phi + theta (the twist step). The crucial detail is that the SAME r and the SAME theta apply for every starting direction phi. A short east arrow and a short north arrow are both lengthened by the identical factor r and both swung through the identical angle theta — the amplitwist plays no favourites among directions.
The phrase to underline is regardless of the arrow's direction. The map does not stretch east-pointing arrows by one factor and north-pointing arrows by another, nor turn them by different angles. Every infinitesimal arrow at z_0, no matter where it points, gets the same r and the same theta. That uniformity is exactly what multiplication by a fixed complex number guarantees, and it is the single feature that a general smooth map of the plane does NOT have — a generic map can stretch different directions differently and shear them apart.
A tiny worked picture
Let us watch the amplitwist on the friendliest example, f(z) = z^2, at the point z_0 = i. Using the power rule d/dz of z^n is n z^(n-1), we get f'(z) = 2z, so f'(i) = 2i. Read off its polar data: |2i| = 2 and arg(2i) = pi/2 (that is 90 degrees). The amplitwist at i is therefore 'double the length, turn a quarter circle'. Every short arrow poking out of i comes back twice as long and rotated 90 degrees counterclockwise.
- Pick a tiny arrow at z_0 = i pointing east, say to z = i + h with h a small positive real number. The arrow is h (length h, angle 0).
- Multiply by f'(i) = 2i: the image arrow is 2i * h = 2h i — length 2h, angle 90 degrees. The east arrow has doubled and rotated to point north.
- Now pick a north-pointing arrow at i, to z = i + i k = (1 + k) i with k small real. The arrow is i k (length k, angle 90 degrees).
- Multiply by 2i: the image is 2i * (i k) = -2k — length 2k, angle 180 degrees. Same doubling, same extra 90-degree turn. The right angle between the east and north arrows is preserved as a right angle.
That last line is the whole point in miniature. The two input arrows started 90 degrees apart; both turned by the same 90 degrees; so they ended 90 degrees apart. The map stretched the local picture (everything doubled) and spun it (everything turned a quarter circle), but it never squashed the angle between two arrows. Stretch-and-spin is a similarity transformation — it changes size and orientation but leaves shapes faithful — which is exactly why infinitesimal squares map to infinitesimal squares, not parallelograms.
Why angles survive: conformality
The miniature example generalizes into one of the headline properties of the subject. Because the amplitwist turns every arrow at z_0 by the same angle arg f'(z_0), the angle BETWEEN any two arrows is left untouched — both arms of the angle rotate together, so their opening stays fixed, and both stretch by the same factor, so nothing distorts. A map that preserves angles between curves like this is called conformal, and what we have just argued is the clean reason a holomorphic function is conformal: angle preservation is a free gift of the rotation-and-scaling structure.
There is an honest fine-print clause, and it matters. Conformality holds only where f'(z_0) is NOT zero. If f'(z_0) = 0 there is no well-defined twist angle (the polar form of 0 has no argument) and no stretch factor that keeps arrows alive — short arrows collapse faster than first order, and angles are generally NOT preserved. Such a point is called a critical point. At a critical point the map typically multiplies angles instead of preserving them: for f(z) = z^2 the derivative f'(0) = 0, and indeed near 0 the squaring map doubles angles (a wedge of opening alpha at the origin opens to 2 alpha), which is the textbook way conformality fails.
The same thing the Cauchy-Riemann equations were saying
The amplitwist is not a new fact bolted on top of the Cauchy-Riemann equations you met two guides ago — it is the SAME fact seen geometrically. Recall that a real-differentiable map of the plane is locally a linear map given by the 2-by-2 Jacobian [u_x, u_y; v_x, v_y], a matrix with four free entries that can stretch, rotate, shear and reflect in arbitrary combinations. The Cauchy-Riemann equations u_x = v_y and u_y = -v_x force that matrix into the special, narrow shape [a, -b; b, a].
Cauchy-Riemann: u_x = v_y, u_y = -v_x
Jacobian = [ u_x u_y ] = [ a -b ] with a = u_x, b = v_x
[ v_x v_y ] [ b a ]
[ a -b ] is EXACTLY 'multiply by a + i b' = r e^(i theta),
[ b a ] a rotation by theta and a scaling by r = sqrt(a^2 + b^2).
and then f'(z) = a + i b = u_x + i v_x.That matrix [a, -b; b, a] is not just any matrix — it is precisely the matrix that acts as 'multiply by the complex number a + i b'. Such a matrix is always a rotation-and-scaling: a rotation by theta = arg(a + i b) composed with a uniform scaling by r = |a + i b|, never a shear and never an angle-distorting stretch. So the Cauchy-Riemann equations, the requirement that f be holomorphic, and the amplitwist picture are three names for one idea: complex differentiability forces the local linear part to be a similarity, a pure stretch-and-turn. The algebra (CR), the linear algebra (the special matrix), and the geometry (the amplitwist) all say the same sentence.
Reading the parts as flow lines
The rotation-and-scaling view also quietly settles a fact this rung keeps circling back to: the real and imaginary parts of a holomorphic function are harmonic conjugates. Because the local map is a similarity that preserves right angles, the level curves of u (where u is constant) and the level curves of v (where v is constant) must cross each other at right angles everywhere f' is nonzero — the conjugate-pair orthogonality you will pin down in the next guide. Pictures of these two families of curves are exactly the equipotential lines and stream lines that fluid dynamics and electrostatics draw.
Notice how much we get from refusing the conjugate. A holomorphic function may use z but never z-bar — the slogan that holomorphy is independence of z-bar. Geometrically, z-bar is the reflection z = x + i y -> x - i y, which flips orientation; an amplitwist may rotate and stretch but it may NOT flip. That is the deep reason conjugation has no complex derivative: a reflection can never be written as a rotation-and-scaling, so it can never be the local action of a holomorphic map. Every property in this rung traces back to that one geometric refusal.