The Rail Thesis

1. A single adoption wave is a sigmoid

Imagine a 10×10 grid of 100 people, all uninfected except one corner. Each cycle, every infected person can pass adoption to their neighbours. Early on almost nobody has it, so growth is slow; in the middle the frontier is wide and growth compounds; at the end there are no new people left to reach, so it flattens. This is logistic growth — the classic S-curve:

2. Chaining waves: the discrete chain

One sigmoid explains one cohort. Bitcoin's history is many cohorts: first cypherpunks, then early internet retail, then exchanges' user bases, then institutional flows, then ETFs. Suppose each new cohort is larger than the last — say the population accessible via each new rail doubles — and suppose each wave only runs until it reaches a small saturation fraction of its cohort before the next rail arrives. Chain them end to end, each wave starting where the previous one levelled off:

Key observation: cutting each wave at any saturation level from 95% down to 2% barely changes the fitted power-law exponent (≈3.1–3.7 across the whole range). What the cutoff changes is the wobble — the steppy remnants of each wave's plateau. A very low cutoff (a few percent) means each cohort is abandoned almost as soon as it starts, and the chain becomes smoother.

3. Why the chain is a power law

Why does chaining exponentials produce a power law (y = A·t^b), not another exponential? Because the doubling of cohort sizes makes each wave contribute roughly a constant multiple to the total, while each wave takes a roughly constant addition of time. Constant-multiple growth per constant-time interval, summed over many intervals, is the discrete skeleton of a power law. On log-log axes the chain is a straight line with slope ≈ b. This is the well-known result that sums of many overlapping exponentials converge to a power law — a pattern seen in relaxation physics, bibliometrics, and network growth. The grid chain is its discrete, visible form.

4. Overlap makes the line straight

The discrete chain still shows stair-steps. Real adoption is not one rail at a time — dozens of rails run simultaneously (many exchanges, many wallets, many countries, ETFs and Lightning side by side), each arriving at its own moment, each growing at its own rate. Simulate many overlapping logistic rails with staggered arrival times and a wide spread of growth rates:

The straightness of the real Bitcoin curves is therefore not a coincidence to be marvelled at — it is a signature. It says the rail arrivals had no characteristic rhythm: no fixed "every N years a new bridge appears." Every era's gaps look statistically like the last era's gaps, magnified.

5. What breaks the straightness: the degradation test

If straightness requires scale-free arrivals, then injecting a characteristic scale should break it. Three injections, one at a time:

All three injections hurt, and the time scale hurts most: rails arriving on a regular schedule roughly quintuple the wobble. Straightness in the real data therefore argues that arrivals had no characteristic rhythm, and that both rail sizes and rail speeds were themselves scale-free — a characteristic size or a characteristic speed leaves a visible mark.

6. From users to price: random wealth draws

So users U(t) ≈ A·t^3.3. Price is a different curve: empirically it grows faster, roughly t⁵. The link proposed here: new adopters are drawn effectively at random from the world population, whose wealth distribution is heavy-tailed (Pareto). Most new users bring little capital; a few bring enormous amounts, and the big ones move price — they buy aggressively and lift the offers that existing holders are willing to accept. As the user base grows, every new cohort samples the wealth tail more deeply.

This step carries an honest negative result: with a realistic wealth tail (α ≈ 1.5, finite mean), random wealth draws reproduce the straightness of the price curve but not its extra steepness. Total demand grows in proportion to users (mean-dominated), so price ≈ users × average wealth. The whale lottery alone does not make price rise faster than users.

7. The rail wealth ladder

The amplifier: successive rails do not open random populations — they open wealthier ones. Cypherpunks had pocket money; internet retail had savings; exchange-verified users had brokerage accounts; institutional rails (and above all ETFs) channel pension funds and family offices. If the wealth scale of the population each rail exposes grows like t^w, then price grows like t^3.3·(1+w):

8. The thesis, assembled

1. Adoption spreads within any connected cohort as an exponential that saturates early — a few percent of the cohort may be enough for the next, larger rail to become worth building (low saturation cutoffs barely change the aggregate).
2. Successive rails open successively larger cohorts. Chained waves of exponential growth sum to a power law, not an exponential.
3. Real rails are many and overlap, arriving at scale-free intervals. The overlap removes the stair-steps; the scale-free arrival pattern is precisely what makes the log-log line unreasonably straight. A characteristic schedule would be visible in the data as curvature — it is not there.
4. New adopters are random draws from a heavy-tailed wealth distribution. Their aggregate demand converts the user power law into a price power law of the same exponent, with the same straightness.
5. Each rail's cohort is wealthier than the last. This wealth ladder multiplies the price exponent by (1 + w) and explains why price (~t⁵) rises faster than users (~t³), i.e. why price per user grows.

9. Practical impact: when the wealth ladder tops out

The model's requirements have opposite signs. The good one: cohorts need not saturate before the next rail fires — a few percent is plenty (Section 2 showed the exponent barely moves between 2% and 95% cutoffs). The binding one: successive cohorts must be wealthier, not just larger. Larger populations exist; wealthier ones are a finite ladder. A plausible reading of the real-world ladder:

Simulate the topping-out directly: the cohort wealth scale grows as t^w until a break era, then freezes. Drag the slider to move the break era and watch both panels. The solid curve is the ladder continuing; the dashed curve tops out.

10. The real data, tested: cycle decay and the break already happening

Sections 1–9 built the mechanism from first principles. This section tests it against the real record — thirteen years of Bitcoin price — and finds three things the paper predicted in advance, plus one that changed our view.

First: cycles decay, measurably. Take the real daily price (2010-07 to 2026-09, 195 monthly closes embedded in this page), fit the power-law trend, and measure each cycle top's overshoot above the trend in log units. Successive tops decay. Fitted to the first four tops alone the decay is remarkably clean — a straight line on a log-amplitude axis (RMSE 0.048), half-life about 8 years. This is the cycle decay the model predicted: each enthusiasm wave moves a smaller fraction of a deeper market.

Second: the trend itself is bent — the break has likely already happened. Fitting a broken power law (old exponent before a break year, new exponent after, continuous at the break) beats a single exponent for every break year from 2020 to 2024. The best-fit post-break exponent lands near 3.1–3.4 — the user-growth floor from Section 9 — and the data cannot yet distinguish break years inside that range. Crucially, the "collapsing cycles" anomaly (the 2025 top looked far below the old decay law) resolves once overshoots are measured against the bent trend: the cycle collapse and the exponent break are the same event seen two ways.