First PrinciplesStart anywhere. Prove it, then move on.

Rung 24

Series and convergence

Infinitely many terms that add to a finite number — sometimes.

Best attempted after 19. Limits. Nothing stops you trying this now — the gate will tell you if you were right.

The gate

Show that a geometric series with ratio below one converges, and give its sum. Then sum the harmonic series in code out to a very large number of terms, show it is still growing, and explain why no amount of computing could settle the question — only a proof can.

Nobody checks this but you. Do it honestly and the rungs above hold; do it loosely and they will not, somewhere further up where the cause is much harder to find.

Add infinitely many positive numbers and you might get a finite answer. Whether you do depends entirely on how fast the terms shrink, and the boundary between the two behaviours is sharper and stranger than intuition suggests.

Why this is on the ladder

Two payoffs. First, the idea that a complicated function can be represented as an infinite sum of simple ones is how transcendental functions are actually computed — every sin your machine evaluates is a truncated series.

Second, and more immediately: this rung is where you learn that a computation can be evidence and never proof. That lesson is the one that protects you at rung 45, when a training run looks like it has converged.

Do this

Geometric first. 1/2 + 1/4 + 1/8 + …. Compute partial sums; watch them approach 1 without reaching it. Then derive the closed form for ratio r below 1, and see the algebra confirm the pattern.

Now the harmonic series: 1 + 1/2 + 1/3 + 1/4 + …. The terms shrink to zero, so it ought to settle. Sum it in code to a million terms, then ten million. It keeps climbing — slowly, by ever less, but it climbs.

It diverges. It never settles, for any bound you name. The grouping proof shows why: terms 3 and 4 together exceed 1/2; terms 5 through 8 together exceed 1/2; each such block adds at least another half, forever.

No computation could have told you that. Your sum reached maybe 16 after ten million terms and looked exhausted. Only the proof settles it, and that gap between "looks converged" and "is convergent" is the whole rung.

Where people get stuck

Believing that terms shrinking to zero is enough. It is necessary and not sufficient, and the harmonic series is the standing counterexample.

The other snag is trusting the partial sums. A slowly diverging series and a slowly converging one look identical on any finite screen — which is exactly the point being made.

Reading