Our laundry recently produced two left socks. Both belonged to my spouse, both were toe socks (the kind knit with a separate pocket for each toe), and each one had a different design. Somewhere between the hamper, the washer, and the dryer, both right socks went missing. What was left was two perfectly good socks that could not be worn together. A left toe sock goes on the left foot the same way a left shoe does. There is no turning it around, and no borrowing another left sock to stand in for the missing right one.

My half of the drawer cannot get into that state, and that is on purpose.

How My Drawer Got Simple

My family owns every kind of sock. Short ones, long ones, thin ones, thick ones, compression socks, soccer socks. I used to have a version of most of those myself: dress socks, cotton socks, long socks for different occasions. Each morning that meant a small search through the drawer for two socks that actually matched, and each laundry day meant sorting them back into pairs.

A few years ago I simplified. I picked one sock (a black ankle sock), bought enough of them, and let the rest go. The idea came from the way Barack Obama described his wardrobe to Michael Lewis in a 2012 Vanity Fair profile: he wore only gray or blue suits because he was "trying to pare down decisions," so he could spend his deciding on things that mattered more. I did not need two colors. I went all in on one.

The result is that my socks are fungible: any one is as good as any other. Any sock goes on either foot, and any two socks make a pair. There is no sorting step after laundry, no search step in the morning, and no such thing as a sock without a match. If one sock gets a hole, I throw away one sock. If the dryer eats one, the drawer has one fewer sock and nothing else changes.

The kids have since joined me. They now follow the same rule I do: one style of sock for school, sports, and everything else. Their socks used to be one more pile to sort after every wash. Now they work the same way mine do, and the laundry basket has one fewer kind of sock to match. It is a simple life, and it got simpler with each person who moved to it.

What a Matched Pair Costs

My spouse's drawer works differently, and it is worth being precise about why, because toe socks are not a mistake. People wear them for real reasons: fit, comfort, keeping toes separated, and, for my spouse, the designs. But every one of those socks carries two constraints at once. It is handed (it only fits one foot) and it is unique (only one other sock in the house matches its design). Put together, that means each sock has exactly one valid match in the whole drawer.

That one fact changes what a loss does:

DrawerValid matches per sockLosing one sock costsLeftover after losing one
Identical, unhanded (mine and the kids')Every other sockOne sockNothing stranded
Unique, handed (my spouse's)Exactly oneA whole pairOne orphan sock

In my drawer, losses are linear: lose three socks and I have three fewer socks. In my spouse's drawer, each loss also strands a survivor. Lose one sock from each of two different pairs and you are holding two orphans, and if both orphans happen to be lefts, they cannot even be pressed into service as an odd pair. That is how we ended up with two left socks: two losses produced zero wearable pairs and two socks that still take up space in the drawer.

The cost also shows up before anything is lost. Every matched pair has to be found again after each wash, so sorting is work proportional to the number of socks. In a drawer of identical socks, that work is zero.

Design for the Spare That Fits

The principle underneath the laundry is one engineers have relearned many times: when a part must match one specific other part, every loss costs double and every repair needs that exact part. When parts are fungible, losses cost only themselves and any spare is a repair.

Manufacturing learned it with interchangeable parts. Before standardized parts, a musket was hand-fitted, and a broken lock could only be repaired by a craftsman fitting a new piece to that particular gun. Once parts were made to a common specification, a field repair meant grabbing any spare from the box.

Connectors learned it more recently. The older USB-A plug (USB stands for Universal Serial Bus) only goes in one way, and most people know the ritual of trying it, flipping it, and trying again. USB-C is symmetric: it has no left or right, and it goes in either way on the first try. The cable gave up nothing to lose its orientation.

Software learned it with servers. A hand-configured machine that only works with one particular database, holding settings nobody wrote down, is a toe sock: when it fails, you need that exact machine back, and there is no spare that fits. A stateless service behind a load balancer is my sock drawer. Any instance can handle any request, a failed instance costs one instance, and the fix is to start another one from the same image.

Fungible is not always the right answer, and shoes show where it stops. People have tried fungible shoes: for long stretches of history, many shoes were made "straight," on a single form, so either shoe fit either foot. It did not stick. A foot is not symmetric, and a shoe is stiff enough that the difference matters, so shoes came to be made as lefts and rights and have stayed that way. The handedness pays for itself in fit. A thin, stretchy sock is different. It conforms to whichever foot it lands on, so making it handed adds a constraint without buying much in return. The complication belongs at the shoe level, not at the sock level.

The same check works on almost any design: if I lose one of these, what else stops working, and can any spare replace it? If losing one piece strands another, the design has built a matched pair. Sometimes that is the right call. Shoes are handed because feet are. A primary database and its replica may need to know about each other. Uniqueness can be the whole point, as it is with my spouse's socks. But the matched pair should be a cost someone chose to pay for a specific benefit, not the default nobody questioned.

Two left socks are what a matched-pair design looks like when it fails: losses that cost double, and leftovers that cannot be used. My drawer of identical black socks is not better taste. It is a system of fungible parts with no pairs to break, so nothing in it can be stranded. The same question applies to the systems I build at work, and it connects to the habit in Minimize Humans as Glue of fixing the design instead of becoming the person who keeps it running: before adding a part that needs one specific mate, ask whether a fungible part would do. Most of the time it would, and then the dryer can eat whatever it wants.