On knowing what to keep.
Aesthetic has come to mean a mood board. A colour, a filter, a vibe. Someone will tell you their brand has a clean aesthetic and mean that it is beige.
Aisthesis is Greek for perception, and when Baumgarten formalised the discipline in 1735 he defined it as the science of sensory knowledge: the claim that the senses are a way of knowing things, with their own rigour, not a decorative layer applied after the thinking is done.
Everything we do rests on that distinction. Aesthetics is not applied at the end, like paint. It is judgement, and judgement can be wrong.
A person whose verdict carries weight because of five things. Strong sense. A delicate sentiment. Improved by practice. Perfected by comparison. Cleared of all prejudice.
Not one of those five is a gift. Every one is a consequence of hours.
Read the list as a job description, because that is what it is. It belongs to Hume, who set out the difficulty in 1757: beauty, he wrote, is “no quality in things themselves: It exists merely in the mind which contemplates them.” Every judgement of taste is subjective. And yet we argue about them as though someone could be wrong, because someone can be. The five conditions were his way out, and what they describe is not a sensibility but a practice.
Taste is not a personality trait we happen to have. It is the residue of having looked at a very large number of things closely and argued about them afterwards. That is why we can charge for it, and why we do not think that is an embarrassing thing to say.
The Western tradition tends to arrive at beauty by addition: proportion, ornament, resolution, the sublime. There is an older answer that goes the other way.
Okakura, writing in 1906, made the Taoist case that the reality of a room is not the walls but the empty space they enclose. The vacuum is the part you use. Japanese practice builds on that inversion: “ma”, the interval deliberately left; “wabi-sabi”, beauty located in the incomplete and the weathered; and the Shinto instinct that beauty is drawn out of a material rather than imposed on it.
The inversion, stated as plainly as it can be stated.
This is the part clients find hardest, and the part that matters most. Most briefs ask for more: more sections, more messages, more proof. Nearly every improvement we actually make is a removal. The gap is not the absence of the design. Frequently it is the design.
A model has never wanted anything, refused anything, or had to live with a decision. It cannot be disappointed in its own work.
We use these tools daily and we are not nervous about them. A generative model is a machine for producing the statistically likely. Trained on everything, asked for something, it returns the centre of the distribution: the average of all the work that came before, rendered competently. That is a real achievement and it is also the exact opposite of taste. Feed the output back into the training data and the average tightens on itself, which is why so much of the web now looks like the same website wearing different logos.
Wittgenstein wrote that ethics and aesthetics are one. He meant that both sit outside the world of facts: they are not things you can be informed about, they are the stance you take toward the whole. A model has no stance.
It can generate. It cannot judge.
There is a category we get mistaken for.
A marketing agency is organised around metrics, and built to be obsolete by the next quarter. That is honest work. It is not this. We are organised around meaning, which is slower and worth more over ten years than any quarter of it. Agencies promote what somebody else built. We are the somebody else.
The difference shows up as a refusal. We will not run the trend, add the section because a competitor has one, or ship the version that tests slightly better and means nothing. Being undefined is not indecision. It is a refusal to be reduced to a formula, because a formula is exactly what a machine can now execute perfectly and for free.
Harmony is the least fashionable word in this essay and the one we would keep if we could only keep one.
It does not mean pleasant. It means the type, the interval, the colour, the pace and the words are all answering to the same idea, and nothing in the thing is arguing with anything else in it. It is audible. People who could not tell you why can tell you when it is absent.
The making is now cheap. The caring whether it coheres is not.
The studio writes a few of these a year. Ask to be sent the next one, and nothing else will ever be sent.
Baumgarten, Aesthetica (1735) · Hume, Of the Standard of Taste (1757) · Kant, Critique of Judgment (1790) · Okakura, The Book of Tea (1906) · Wittgenstein, Tractatus (1921)