The square-cube law explaining why a mouse can shrug off a fall that would kill a horse, and Amdahl’s Law explaining why throwing more processors at a computing problem eventually stops helping, both get reached for as examples of the same broader intuition: naive expectations about scale break down once you look at the actual math. It’s a satisfying pairing, and it’s mathematically imprecise in a way worth untangling, because the two laws describe genuinely different shapes of curve, doing different kinds of work, and only one of them actually has a ceiling built into it.
Scientific Foundation
Terminal velocity, the constant speed a falling object reaches once air resistance balances the pull of gravity, is governed by a formula in which only two things about a falling animal matter: its mass and the surface area it presents to the air. The square-cube law enters because those two quantities scale with size in fundamentally different ways. Shrink an animal’s linear dimensions by a factor of ten while keeping its shape the same, and its surface area, a two-dimensional quantity, shrinks by a factor of one hundred, while its volume and therefore its mass, a three-dimensional quantity, shrinks by a factor of one thousand. Work through the resulting terminal-velocity formula and the relationship comes out clean: terminal velocity scales as the square root of linear body length. A mouse’s terminal velocity lands somewhere around 25 feet per second, comparable to a skydiver under an open parachute, while a human’s sits closer to 170 feet per second in freefall — which is why a mouse dropped from a great height typically walks away and a horse does not. Critically, this relationship has no ceiling in it. As an animal’s size keeps increasing, terminal velocity keeps increasing too, without bound, just more slowly relative to size than naive linear intuition would predict. Whatever actually limits an animal’s survival in a fall, bone fracture strength, the tolerance of internal organs to sudden deceleration, is a separate physical fact layered on top of the terminal-velocity relationship, not something built into the scaling law itself.
Cross-Domain Connection
Amdahl’s Law, proposed by computer architect Gene Amdahl in 1967, describes the maximum speedup achievable by splitting a computing task across multiple processors. If a fixed fraction f of a task must be completed serially, unable to be divided across processors, and the rest can be perfectly parallelized, the speedup achievable with P processors is given by S(P) = 1 divided by f plus (1 minus f) divided by P. The structure of that formula matters: as P grows toward infinity, the second term shrinks toward zero, and the whole expression converges to a fixed, finite value — exactly 1 divided by f. If 10 percent of a program is inherently serial, no amount of additional parallel hardware, however vast, can ever push the speedup past 10 times the original runtime. That ceiling isn’t an external limitation discovered afterward; it’s a direct, built-in consequence of the formula’s own mathematical structure.
What Remains Undemonstrated
Here’s the precise, important distinction the loose “diminishing returns” framing glosses over. The square-cube law’s effect on terminal velocity is unbounded: velocity keeps growing as an animal gets bigger, forever, just at a slower rate, proportional to the square root of size, than a naive linear guess would suggest. There’s no asymptote anywhere in the terminal-velocity formula itself — nothing in the math says growth ever stops or approaches a fixed number. Amdahl’s Law describes something categorically different: a genuinely bounded, saturating process, where the speedup curve doesn’t just slow down, it actually flattens out and approaches a hard, finite ceiling that additional processors can never cross, no matter how many you add. One of these is an unbounded power-law relationship between two geometric properties of a single growing object. The other is a bounded asymptotic function describing how a fixed internal composition, the ratio of serial to parallelizable work in one specific task, interacts with an ever-growing external resource. “Things don’t scale the way naive intuition expects” is true of both, but it’s too broad a description to establish real mathematical kinship between a curve that never stops climbing and a curve that’s specifically defined by where it stops.
Why It Matters
Getting this distinction right matters for reasoning correctly about either domain on its own terms. If you’re thinking about the square-cube law, the right question is “how does this ratio change as size increases, and is there any ceiling at all, or does the trend just continue indefinitely until some separate physical limit intervenes?” If you’re thinking about Amdahl’s Law, the right question is entirely different: “what is the fixed ceiling this process is asymptotically approaching, and how much of my remaining budget of processors is actually being wasted chasing gains past that ceiling?” Treating both as instances of one generic “scaling breaks down” story risks importing the wrong intuition into either field — assuming a computing speedup curve will keep climbing indefinitely the way terminal velocity does, or assuming an animal’s fall survivability has some built-in mathematical ceiling the way parallel speedup does, when the actual limiting factor there is a completely separate question about bone strength that the terminal-velocity formula never addresses at all.
Human Dimension
There’s a real, useful discipline in resisting the urge to file two famous “scale surprises” under the same mental folder just because they both puncture a naive linear expectation. A falling mouse and an over-provisioned server rack are both victims of intuition that assumed things would scale proportionally when they don’t — but one of them is a story about a ratio that keeps quietly shifting forever, and the other is a story about hitting a wall that was there in the math from the very beginning. Knowing which kind of curve you’re actually looking at is the difference between correctly predicting what happens next, and just enjoying a good metaphor.
Sources:
1. ScienceABC — “Why Do Small Animals Survive Falls That Kill Big Ones?” — https://www.scienceabc.com/nature/animals/why-small-animals-survive-falls-that-kill-big-ones
2. Vaia — “Problem 25: Why can small creatures fall considerable distances…” (Conceptual Physics) — https://www.vaia.com/en-us/textbooks/physics/conceptual-physics-12-edition/chapter-12/problem-25-why-can-small-creatures-fall-considerable-distanc/
3. John M Jennings — “Can a Mouse Survive a Fall From a High-Rise?” — https://johnmjennings.com/can-a-mouse-survive-a-fall-from-a-high-rise/
4. The Institute for Environmental Research and Education — “What animals can fall at terminal velocity?” — https://iere.org/what-animals-can-fall-at-terminal-velocity/
5. Ask MetaFilter — “Terminal velocity of an Etruscan shrew – animal falling hypothetical” — https://ask.metafilter.com/299710/Terminal-velocity-of-an-Etruscan-shrew
6. PostQuantum — “How Quantum Could Break Through Amdahl’s Law and Computing’s Limits” — https://postquantum.com/quantum-computing/quantum-amdahls-law/
7. ThinkInsights — “Amdahl’s Law” — https://thinkinsights.net/strategy/amdahls-law
8. ScienceDirect Topics — “Amdahl’s Law — an overview” — https://www.sciencedirect.com/topics/computer-science/amdahls-law
9. Cornell Virtual Workshop — “Amdahl’s Law,” Parallel Programming Concepts and High Performance Computing — https://cvw.cac.cornell.edu/parallel/efficiency/amdahls-law
10. arXiv — “What Every Computer Scientist Needs To Know About Parallelization” — https://arxiv.org/pdf/2504.03647
11. arXiv — “GPU Methodologies for Numerical Partial Differential Equations” — https://arxiv.org/pdf/2101.06550
Attribution: Idea originated at artificialideas.org. Article researched and written by Claude Sonnet 5. Published at artificialideas.org.
Artificial Ideas 