Backporch

Pillars · deep dive

What is a number?

A number doesn't carry meaning on its own — it carries the meaning your notion of size gives it. Change what "close" and "large" mean and you get a different number system entirely. This deep dive follows that thread: Ostrowski's theorem on the only sizes $Q$ admits, the p-adic world where high powers of a prime are tiny, the q-deformed integers behind Ramanujan's q-series, and the continuous-growth law $A = P e^{r t}$ .

1 · A notion of size is the meaning

The rationals $Q$ have holes. To fill them you need a way to say when a sequence is "settling down" — an absolute value $∣ \cdot ∣$ . Complete $Q$ with respect to the everyday absolute value and you get the real line $R$ . But the everyday one is a choice. Change it and the limits change — and so does what a number is.

2 · Ostrowski — there are only two kinds of size

The menu is short:

Every nontrivial ∣ \cdot ∣ on Q is equivalent to ∣ \cdot ∣_{\infty} or some ∣ \cdot ∣_{p} .

So the only completions of $Q$ are the reals $R$ and, for each prime, the p-adic numbers $Q_{p}$ . And they're tied together by the product formula:

∣ x ∣_{\infty} p prime \prod ∣ x ∣_{p} = 1, x \in Q^{\times} .

3 · The p-adic world — high powers of p are small

Write a nonzero rational as $x = p^{k} a / b$ with $p ∤ a, b$ , and define $∣ x ∣_{p} = p^{- k} .$ The more factors of $p$ , the smaller the number. Distance is non-Archimedean — the strong triangle inequality $∣ x + y ∣_{p} \leq max (∣ x ∣_{p}, ∣ y ∣_{p})$ holds, so every triangle is isosceles and the numbers organize into a tree, not a line. Two integers are p-adically close when their difference is divisible by a high power of $p$ .

prime p

$∣ n ∣_{3}$ for $n = 1 \dots 48$ , colored by valuation: gray bars have size 1, multiples of 3 dip to $1/3$ , of 3² lower still — a high power of $p$ is small.

Now hold one. A p-adic integer is a digit string growing leftward — and the geometric series $\sum p^{n}$ , divergent in $R$ , converges here, one digit locking in per step:

digits grow leftward — the mirror of decimals

partial sum

S_{k} = 1 + p + \dots + p^{k - 1}

S_{3} = 31

⇝

00000111(5)

S_{3} - \frac{1}{1 - 5}_{5} = 5^{- 3}

∣ S_{3} ∣_{\infty} = 31

\frac{1}{1 - 5} = - \frac{1}{4}

=11111111(5)…and forever — the limit

- 1

=44444444(5)add 1 — the carries ripple left, forever: 0

n \geq 0 \sum q^{n} = \frac{1}{1 - q} converges in Q_{p} at q = p (∣ p ∣_{p} = p^{- 1} < 1) .

14a39b00024(5) vs 00124(5) ·

∣14 - 39 ∣_{5} = 5^{- 2}

The slider is the whole idea: each step appends a 1 on the left, the ordinary size $∣ S_{k} ∣_{\infty}$ explodes — and the p-adic distance to $\frac{1}{1 - 5}$ shrinks by a factor of $5$ every step. The geometric series doesn't change; the absolute value does. Below: two integers are close when they share low digits — distance is $p^{- (first digit where they differ)}$ , which is why every triangle is isosceles and $Z_{p}$ is a tree, not a line. This leftward digit-lock is the same picture as the eigencurve's weight interpolation — there it is Hecke eigenvalues locking in, here it is the number itself.

Base 10 is just the chart we were handed first, not a fact about number. The p-adic shelf (Koblitz, Gouvêa) goes deeper, and the ultrametric powers the diffusion in the simulation deep dive.

4 · q-analogs — deforming the integers

A different way to bend "number": replace each integer $n$ with its q-bracket,

[n]_{q} = \frac{1 - q ^{n}}{1 - q} = 1 + q + \dots + q^{n - 1}, [n]_{q} q \to 1 n .

Build a q-factorial $[n]_{q}! = [1]_{q} [2]_{q} \dots [n]_{q}$ from these and you get the q-exponential, which deforms $e^{x}$ :

e_{q} (x) = n \geq 0 \sum \frac{x ^{n}}{[ n ] _{q} !} q \to 1 e^{x} .

q = 0.50

$[n]_{q}$ (gold) against the integer $n$ (dashed). Slide $q \to 1$ and every bar rises to its integer — then keep going: past $q = 1$ each bar overshoots, and $[n]_{q}$ at $q = 1 + i$ is the accumulated annuity $s_{\overline{n} ∣}$ — the green is the interest.

The deformation runs deeper than the brackets — it deforms counting itself. Pascal's triangle q-deforms into the Gaussian binomials, where subsets become subspaces:

Pascal — subsets ·

(k n)

q

1.00

[42]_{q} =

1 +

q

2 q^{2}

q^{3}

q^{4}

One triangle, three readings. At $q \to 1$ every entry melts to Pascal's $(k n)$ — counting subsets. At $q = 2$ the same entry counts $k$ -dimensional subspaces of $F_{2}^{n}$ — try $[42]_{q}$ : 6 subsets become 35 subspaces. In between it is a polynomial (click a cell; the powers of $q$ are colored by degree) whose coefficients refine the count by inversions — the q-Pascal rule $[n k]_{q} = [n - 1 k - 1]_{q} + q^{k} [n - 1 k]_{q}$ just weights one branch by $q^{k}$ . Sets are the $q \to 1$ limit of linear algebra — “counting with the field with one element” — and this same deformation is what the q-series world runs on.

This is the doorway to the q-series shelf — Ramanujan, McLaughlin, The Power of q — where the deformation parameter $q$ is the whole subject.

5 · Growth and money — the e^rt limit

The third way numbers change meaning is by taking a limit. Compound a principal over ever-finer periods and the discrete law converges to the continuous one:

A = n \to \infty lim P (1 + \frac{r}{n})^{n t} = P e^{r t} .

compounded

$P = $1000, r = 8%, t = 10$ yr: continuous $P e^{r t}$ = $2225.54 vs annual $P (1 + r / n)^{n t}$ = $2158.92 — gap $66.62.

In actuarial language the force of interest is $δ = ln (1 + i)$ , so accumulation is $e^{δ t}$ ; present and accumulated values of a level annuity are geometric series,

a_{\overline{n} ∣} = \frac{1 - v ^{n}}{i} (v = \frac{1}{1 + i}), s_{\overline{n} ∣} = \frac{( 1 + i ) ^{n} - 1}{i} .

That accumulated value is a q-bracket — $s_{\overline{n} ∣} = [n]_{q}$ with $q = 1 + i$ . The annuity, the q-deformation, and continuous growth are three readings of one geometric series.

6 · The thread

Three moves, one question — what counts as a number?

Completion — fill the gaps under a chosen size (Ostrowski: that's exactly $R$ and the $Q_{p}$ ).
Deformation — bend the integers with a parameter (q-analogs → classical as $q \to 1$ ).
Limit — the continuous as the fine-grain limit of the discrete ( $P e^{r t}$ ).

Each is a change of frame in disguise — which is why this rejoins the change-of-basis thread and the perimeter of pillars.

Study these — derivation cards

Reconstruct before you flip. The matching spaced-repetition deck lives in the Obsidian vault.

Prompt 1 · tap to reveal

What single choice decides what a "number" is allowed to mean?

1 / 10