Specimen Cabinet · Est. 1300–2006

The Cabinet of
Named Laws

Thirty-nine named laws, cognitive biases, and eponymous principles — from a razor sharpened seven centuries before the internet to a law of ambiguity coined in a forum post. A field guide that keeps the difference between a joke and a theorem visible, rather than flattening everything into the same wise-old-saying register.

39Specimens
706Years spanned
5Classifications
1Actual theorem
Enter the archive
✦

A grab-bag of folk wisdom that shares one surface feature — a person's name attached to a pithy rule — and almost nothing else underneath.

Some started as jokes and turned out to be true. Some are solid, tested facts about how money and numbers behave. Some are ways of thinking far older than the person they're named after. And some were born on the internet, barely twenty years ago.

This cabinet's job is to be honest about the difference: a joke isn't the same as a proven fact, and it's worth keeping them apart. Every entry is labelled with what it actually is.

I

Cabinet I · Formal & Proven

The three with actual proofs

Unlike almost everything else here, these three aren't just clever sayings — they're backed by real maths you can actually see. So instead of quoting the equations, we drew them: a shape you can spin, a chart, and a web of connections you can play with.

CAP Theorem

Accuracy · Always-answering · Surviving a split — pick two

Proposed in 2000 and actually proven true in 2002 — one of the rare things here that's a hard fact, not a saying. If you spread your data across several machines, you can't have everything at once.

When those machines lose contact with each other, you're forced to choose: give an answer that might be out of date, or refuse to answer until they reconnect. Drag the shape to turn it around.

SPEC · CAP-2002 Rotatable proof · drag to orbit
SPEC · METCALFE-1983
6 · 15
Every dot connects to every other dot

Metcalfe's Law

A little more people → a lot more connections

From the co-inventor of Ethernet, back in 1983. A network gets far more valuable as more people join — because each new person can connect to everyone already there, and those connections pile up fast.

One honest catch: even Metcalfe later admitted the real effect isn't quite as strong as the snappy version suggests. Slide the node count up and watch how quickly the lines multiply.

Amdahl's Law

Fix the big slow part, not the small easy one

From 1967. Speeding up a small piece of a job barely helps the whole thing — even if you make that piece instant. What matters is how big the part you fix is.

The eye-opener: take a slice that's only a tenth of the work and make it 10,000 times faster, and the whole job still only gets about 1.1× quicker. Each line below flattens out at its own ceiling — no matter how many machines you throw at it.

SPEC · AMDAHL-1967 More machines → speed gain flattens out
II

Cabinet II · Software Engineering Folklore

All non-trivial abstractions leak

From a 2002 essay by Joel Spolsky: tools that hide messy details to make life easier are great — right up until they break, and suddenly you have to understand the thing underneath anyway. Below: the layers of the internet, stacked up, with light leaking through every gap where one layer doesn't quite line up with the next.

SPEC · LEAKY-2002 Each layer quietly leans on the one below it

The layer that gives itself away

The internet stacks simple layers on top of each other so each one can promise something the layer below doesn't. It works beautifully — until a cable gets cut somewhere, and suddenly you're forced to understand all the messy plumbing you thought you could ignore.

…which leads straight to Hyrum's Law

Those little leaks are how people first notice the hidden details of how something works. And once enough people notice, somebody starts relying on them — even the accidental quirks and bugs. That's where the phrase "bug-for-bug compatible" comes from.

The Long View · 1300 → 2005

Seven centuries, ranked not scaled

A linear year-axis would crush thirty of these into an unreadable knot around 1948–2005 while three medieval outliers float alone. So the cabinet spaces specimens by chronological rank — evenly, one per slot — printing the true year on each plaque and grouping by era. Older entries recede into fog; recent ones stand sharp and near.

→ Scroll the rail sideways · hover any plaque to bring it forward out of the fog

III

Cabinet III · Cognitive Biases

The chart that argues with itself

Dunning-Kruger is the most famous bias here — and the most argued-over. Instead of drawing the well-known curve as if it's settled fact, this one shows the honest doubt: a lot of that "clueless and confident" pattern might just be a quirk of the maths, not people being genuinely oblivious.

Being sure isn't being right

The famous 1999 study: the people who did worst on a test guessed they'd done far better than they had. The popular takeaway — if you're bad at something, you're too bad to even realise it.

But later researchers showed you can recreate almost the same graph from plain statistical flukes, without anyone actually being clueless. So the strong version of the story is shakier than it sounds.

The useful takeaway either way: don't trust anyone's confidence — including an AI's — as proof they're right. Feeling sure and actually being correct barely go together.

SPEC · DUNNING-KRUGER-1999 What people think of themselves vs how they really did
IV

Cabinet IV · The Statistical Wing

Laws you can watch happening

Three of the statistical specimens aren't just sayings — they're mechanisms. Each model below actually runs: set a target and watch the number corrupt, stretch a deadline and watch the work swell to fill it, and trace the tiny slice of causes that does most of the work.

Goodhart's Law

Aim at the measurement and it stops measuring

Two lines start out telling the same story: the number you track, and the real thing it stands for. Then someone turns the number into a target.

From that moment the lines part ways — the metric soars while the real thing stalls and sags. Drag the flag to choose when the target is set; the corruption starts exactly there, every time.

SPEC · GOODHART-1975 Drag the flag — the gap opens where the target lands
SPEC · PARKINSON-1955
6 days
Same job, any box — the box always ends up full

Parkinson's Law

Work expands to fill the time available

The job below genuinely needs about a day and a half. Give it more days and the job doesn't get bigger — the faffing does: warming up, tinkering, polishing detours that feel like progress.

And notice where the real work sits: crammed against the deadline, every single time. Slide the deadline out and watch it happen.

Pareto Principle

The vital few carry the trivial many

Line up a hundred causes from biggest to smallest and add up what they produce. The curve leaps at first — the handful that matter — then crawls through the long tail.

The ring marks the famous corner: 20% of causes, ~80% of results. Sturgeon's cut lives on the same curve — the first tenth that isn't rubbish. Run your finger along it.

SPEC · PARETO-1906 Hover the curve — top share of causes → share of results
V

Cabinet V · The Connections

No law stands alone

Filed in separate drawers, the specimens keep talking to each other — razors sharpen razors, the metric-worship laws corrupt together, and the internet-native rules huddle in their own corner. Every thread below is a real relationship, drawn by hand. Hover a specimen to see its neighbours; click one to pull open its drawer.

SPEC · INDEX-ALL 39 specimens · hand-drawn threads · click any node to open its drawer

The Full Collection · 39 Specimens

Every drawer in the cabinet

The complete field guide. Each specimen carries its classification, origin, exact quote, an honest test against real solo-operator work — and now the threads to its neighbours. Search it, sort it, or pull a drawer at random. Click a catalogue number to copy a direct link to that specimen.

⌕
Showing 39 / 39 specimens