Why time speeds up — and what a 1937 puzzle has to do with it

Why Time Speeds Up — and What a 1937 Puzzle Has To Do With It

CTF / PLCT · Research Notes

Why time speeds up — and what a 1937 puzzle has to do with it

A funnel that narrows, a proof two lines long, and a ratio that keeps coming out 2 : 1 no matter where you look.

Ask anyone over forty whether the years are moving faster. The answer is almost always yes, and it arrives immediately. The usual explanations — fading memory, less novelty, a year being a smaller slice of a longer life — are all real. But they’re downstream of a question nobody asks: is there a structural reason time should compress? Not biological. Not psychological. Arithmetic.

This is a note about two results that were developed separately and turned out to be about the same thing. One is a curve. The other is a fact about one of the most famous unsolved problems in mathematics. Neither needs the other to stand up, which is why it’s interesting that they agree.

The shape

The funnel isn’t a metaphor

Start with a single principle: whatever governs how intervals of time scale should depend only on the rate at which that scaling changes, not on any absolute value. Write that down as an action and solve it, and exactly one curve comes out.

λ(t) = e−βf₀t the unique extremal — the shape is forced, not fitted

λ is how much lived time you get per tick of clock time. It shrinks. Always, from the first moment to the last. That isn’t a poetic description of time speeding up — it’s the definition of it. Nobody chose the shape; it’s the only one the equation permits.

t = 0 t → ∞ λ = 1 λ → 0 the wide end · the past the narrow end · ahead

There’s a second channel most people miss. Conserve angular momentum inside the funnel and the spin accelerates at exactly twice the rate the funnel narrows — and the lived content of one completed turn falls as the cube. Which matters, because what people actually report isn’t “seconds feel short.” It’s wasn’t it just Christmas? Birthdays are angular. Recurrence compresses fastest of all.

A luminous funnel of cyan and magenta filaments narrowing to a bright throat.
The funnel, animated from the equations. The wall lines follow the closed-form spin-up; the coloured meridians are the three families of numbers described below. Radius falls at exponent 1, rotation rises at exponent 2, content per turn falls at exponent 3.
The three families

Every number is one of three things

Work with remainders after dividing by 9 and the numbers split into three groups that behave completely differently. This isn’t a classification anyone imposed — it falls out of the algebra.

0 1 2 3 4 5 6 7 8 0, 3, 6 — no inverse exists 1, 4, 7 — the perfect squares 2, 5, 8 — what squaring can’t reach

The cyan set {1, 4, 7} are exactly the squares. Square any number in this system and you land here, every time. It’s the reversible half — anything that undoes itself lives in cyan.

The magenta set {2, 5, 8} is everything squaring can never reach. You cannot arrive here by any operation that is its own mirror. It’s the one-way half.

The dark set {0, 3, 6} can’t be undone at all — these have no multiplicative inverse. Nothing can act on them and nothing escapes. They’re the still points.

Irreversibility isn’t an interpretation laid over the magenta set. It’s the definition of it. And the same split appears at every scale — modulo 9, modulo 27, modulo 81, all the way up. It’s really a fact about remainders after dividing by 3.

The puzzle

A rule a child can follow that nobody can solve

Pick a number. If it’s even, halve it. If it’s odd, triple it and add one. Repeat.

Every number anyone has ever tested crashes down to 1. Nobody can prove they all do. It has been open since 1937, and Erdős reportedly said mathematics isn’t ready for it. Ignore the halvings for a moment and look only at the odd numbers a trajectory passes through — the spine of the journey. Colour each by its family.

The result

Which colour you land on has nothing to do with where you are. It depends only on how many times you could halve on that step.

3n + 1 ≡ 1  (mod 3)  ·  2 ≡ −1  (mod 3)
⇒  (3n+1)/2k ≡ (−1)k halve an odd number of times → magenta · an even number of times → cyan

That’s the entire proof. Tripling always lands you in the cyan set — always, no exceptions, because 3n+1 can never be a multiple of 3. Then dividing by 2 flips you between the two halves, and dividing k times flips you k times. The number you started with cancels out completely.

any odd n any family 3n+1 lands in cyan always · no exceptions ÷2×k k even → cyan happens 1/3 of the time k odd → magenta happens 2/3 of the time

Now count. Halving once happens half the time. Three times, one in eight. Five times, one in thirty-two. Add up every odd case:

½ + ⅛ + 1/32 + … = 2/3 magenta appears exactly twice as often as cyan · in every trajectory · forever
The sharp part

It has no memory

Because the next colour depends only on k, and k is drawn fresh at every step, the previous colour tells you nothing about the next one. The sequence isn’t a pattern with a 2:1 bias. It’s independent coin flips, weighted 2:1.

We tested that on 640,000 consecutive pairs from real trajectories. The probability of magenta after cyan and the probability of magenta after magenta differ by 0.00009 — seven hundredths of a standard deviation from perfect independence.

QuantityPredictedMeasured
magenta share0.666670.66609
P(magenta | previous cyan)0.666670.66621
P(magenta | previous magenta)0.666670.66612
multiples of 3 encountered00
sample size—680,000 steps

Run it on your own number

This computes the real odd backbone in your browser. Every tile is one step, coloured by family. Watch the ratio settle.

magenta0
cyan0
ratio—
multiples of 30

Three more things fall out

The exact long-run frequencies of the individual remainders have denominator 63 — and 63 isn’t arbitrary. It’s 2⁶ − 1, where 6 is how long it takes powers of 2 to cycle around modulo 9. Change the modulus and the denominator changes with it, always as 2φ − 1.

Doubling carries exactly twice the mass: whatever weight a cyan remainder has, its double in the magenta set has precisely double. The 2:1 holds class by class, not just in aggregate.

And it needs the 3. Try 5n+1, or 7n+1, and the structure collapses — you start landing on multiples of 3, the collapse fails, and the ratio isn’t 2. The multiplier that makes the whole thing work is the same number that generates the family where nothing moves.

The picture

Why the coral grows straight

A branching coral of magenta and cyan filaments growing upward from a single bright trunk.
The Collatz coral. Every branch is one number’s journey, drawn backwards so all paths begin at 1. Nothing here is stylised — the turn at each step is the halving count, and the colour is the family.

Each step turns left when k is odd and right when it’s even. Left turns are twice as common — so we drew right turns twice as wide. Two thirds times one, minus one third times two, is zero. The drift cancels exactly, and the coral stands up instead of curling.

That’s worth sitting with. If the ratio were 3:1 or 4:1, this structure would coil into a spiral and fall out of frame. The fact that it grows straight is the theorem, rendered as a shape rather than asserted as a number.

Hundreds of Collatz trajectories spiralling inward to a bright centre.
The same trajectories in polar form. Radius is the logarithm of the current value, so every path winds inward and lands at 1. Two thirds of every journey is magenta.
Where they meet

The arrow is made by running

Here’s the part that made both of these worth writing up together.

Count the lattice at rest and it is perfectly even-handed: twenty-four cyan and twenty-four magenta across the full unit group, thirty-six and thirty-six across the natural working domain. No preferred direction anywhere in the standing structure.

The 2:1 appears only when something iterates. Only when the map runs.

structure chooses nothing · running chooses everything

In June 2026 a physicist at Birmingham built a closed quantum system — a cloud of atoms chilled to near absolute zero, split in two, with one half deliberately ignored — and constructed a clock out of nothing but the entropy flowing between the halves. The clock ordered events correctly. Its rate rose when exchange was heavy, fell when exchange slowed, and stopped entirely when the two halves reached equilibrium.

A residue lattice and a thermodynamic entropy budget, arriving from unrelated directions at the same statement: no flow, no arrow. Time’s direction isn’t a property of the world’s furniture. It’s a property of the world working.

What this does and doesn’t show The cold-atom clock oscillates — it speeds up and slows down as atoms slosh back and forth. The funnel decays monotonically. So the experiment supports the idea that a system’s time rate can be set by its own internal state. It does not confirm the funnel’s direction, and nothing in it touches the arithmetic.
Beyond what’s proved

What this might mean

Speculation — clearly marked

The framework holds that roughly two ninths of processes stay coherent while the rest drift into arithmetic incoherence — still happening in clock time, no longer registering in lived time. The Birmingham experiment suggests a reading for that discarded portion: it’s a dark sector. The degrees of freedom the coherent part stopped tracking. And under the experiment’s own logic, that untracked portion isn’t merely lost — it’s the source of the time the tracked portion experiences.

If that correspondence is real, then forgetting isn’t a failure of the system. It’s the engine.

Speculation — clearly marked

The same curve can be read two ways. Along one life it says lived time compresses as you go. Across lives it says something else: anyone starting later starts further down the funnel, at every age, not just at the end. The offset would be a constant multiplicative factor — uniform across a whole lifespan, not something that accumulates.

That shape is derivable. The magnitude is not, and the gap between a millisecond-scale oscillation and a human lifetime is not closed. Whether anything measurable shows such an offset is an open empirical question, and it is being argued separately rather than smuggled in here.

Speculation — clearly marked

The one loop the Collatz map is known to have — 1 → 4 → 1 — halves twice. An even number. So it never leaves the reversible half. Every journey spends two thirds of its steps in the one-way sector and then terminates in the sector where nothing moves.

Which is what a fixed point ought to look like, in a picture where time is identified with motion that can’t be undone.

Honesty

What would kill this

The arithmetic is checkable and I’d rather it were checked than admired. Four ways it could fail:

Find dependence. If the colour of one step predicts the next in any measurable way, the memoryless result is wrong. This is a one-line test on any trajectory data.

Find a multiple of 3 anywhere in an odd backbone after the first step. One would do it.

Break the ratio. A recursive process built the same way that comes out at 4:1 instead of 2:1, or with even weight inside the families, falsifies the account.

Break the exponents. The funnel predicts that anything it governs compresses cyclically at exactly twice the linear exponent, and delivers content per cycle at three times. A fitted ratio that isn’t 2 kills the two-channel coupling. That test can be run today, in a cold-atom lab, on geometry alone.

A warning to anyone replicating My first two sampling designs rejected a true theorem at five and six standard deviations. Requiring long trajectories quietly selects for small halving counts — and one is odd. Counting complete trajectories re-counts the same final descent thousands of times. Fix the step index instead, start from numbers around 10⁴⁰ so nothing terminates inside your window, and read the residue there. Then everything lands.

The funnel narrows. The coral stands straight. Two thirds of every journey is one-way, and then it stops.

None of this proves that time is a lattice. What it shows is smaller and more specific: the shape of the geodesic is formally identical to the structure of time speeding up, and the arithmetic beneath it produces a temporal asymmetry out of a structure that has none at rest. The asymmetry arrives with the iteration. It is not in the furniture.

The framework began with a feeling, in a hospital bed, in early 2025 — that time flows through a shape, wide behind and narrow ahead. The mathematics came later and by a completely different road. They arrived at the same place.

Griff Gurwell — CTF / PLCT Framework Research Series

Papers, proofs and source code: ctftheory.com
Preprints: Zenodo community “CTF Theory”
Every figure and animation on this page is generated from the trajectories themselves. Nothing is stylised into agreement.




Scroll to Top