01 Jun The Algorithm Has a Wardrobe

3 min read

There is a specific experience that has become common enough to deserve a name. You are watching a video, on TikTok, on Instagram Reels, on YouTube Shorts, and the person on screen is wearing something you have seen before. Not the exact garment. The combination. The specific pairing of wide-leg trousers and a fitted crewneck and loafers that you saw on someone else yesterday, and someone else the day before.

You scroll. The next video: same combination, different colourway. The one after: a variation. Within a week you have seen thirty versions of the same outfit on thirty different people in thirty different cities, and somehow it still gets described as “personal style.”

This is not a trend. A trend has a point of origin, a designer, a subculture, a specific time and place where something started. What you are watching is something different: an optimisation. The algorithm found an aesthetic that performed, and it is showing it to everyone until everyone replicates it, and then it will find the next one.

The Mechanics

Recommendation systems are not neutral surfaces. They have objectives, watch time, engagement, return visits, and they learn, very quickly, which content achieves those objectives. An outfit that photographs well, that reads clearly at small scale, that signals a legible aesthetic category, performs better than one that is genuinely strange or personal or hard to classify. The algorithm doesn’t penalise interesting clothes. It simply surfaces predictable ones more reliably.

The result is a feedback loop. Aesthetics that perform get shown to more people. More people adopt them. More content featuring those aesthetics gets created. More of that content performs. The loop closes. The aesthetic spreads until it is no longer an aesthetic, it is wallpaper.

TikTok named this process, accidentally, by inventing the “-core” suffix. Cottagecore. Darkacademia. Quiet luxury. Old money. Each one a legible, searchable, hashtaggable category that the algorithm can cluster and the creator economy can produce content around. The naming was a service to the machine. Once an aesthetic has a name, it can be indexed. Once it can be indexed, it can be amplified. Once it can be amplified, it can be exhausted, and then replaced by the next one.

The cycle that used to take a decade now takes eighteen months.

What Was Lost

Getting dressed used to require exposure to friction. You lived somewhere, you knew certain people, you had access to certain shops and magazines and subcultures. Your style was formed by what you could actually encounter, which meant it was formed by geography, by class, by accident. These constraints were limiting. They were also generative. People dressed differently in different cities because they were exposed to different things.

The internet dissolved this. A teenager in Lisbon and a teenager in Seoul and a teenager in São Paulo are now watching the same content, exposed to the same recommendation outputs, buying from the same international platforms. The friction is gone. So is the differentiation it produced.

What you are left with is a global aesthetic consensus that updates seasonally and is mistaken, repeatedly, for individual taste.

The Counter-Position

The brands that have understood this, and profited from it, are the ones that refused to be indexable. Margiela doesn’t have an aesthetic that photographs cleanly. Neither does early Helmut Lang, or Yohji Yamamoto, or anything designed by someone who was more interested in the problem of the body than in the problem of the feed. These are clothes that require context to read. They resist the thumbnail.

This is not nostalgia for difficulty. It is an observation about what happens to an aesthetic when it becomes frictionless: it spreads everywhere, and then it means nothing.

Personal style is not what the algorithm recommended to you at scale. It is what you chose when the recommendation wasn’t available. The question is whether enough people remember what that felt like to want it back.

The edit starts now.