# Auto-generating auto-generative systems

Canonical URL: https://jeanyoon.ch/oi/texts/auto-generating-auto-generative-systems
Type: essay / research note
Author: Jeanyoon Choi (최정윤)
Date: 2025-03-10

## Summary

A 2025 manifesto-style text by Jeanyoon Choi that moves from 'a system that generates' to 'the generation of systems that generate'. In a deliberately exaggerated contrast, the twentieth century made the human part of the factory, while AI can put factories, and factory-generating algorithms, in individual hands. It calls for generating alternative housing, interfaces, products and organisations rather than options within fixed terms, while warning that hyper-customisation can conceal a new standardisation: one should ask whether one is customising the system or being customised by it.

## 요약 (Korean)

최정윤은 테일러주의 이후 인간이 시스템의 부품이었다면, AI 시대에는 개인이 '생성하는 시스템을 생성하는' 주체가 될 수 있다고 주장하며, 초개인화가 또 다른 표준화가 될 위험도 경고한다.

## Concepts defined

- **Auto-generating auto-generative systems**: Choi's phrase for automatically generating systems that themselves generate, recursively, so that individuals produce the machinery of production rather than serving it.

## Questions this answers

### What does 'auto-generating auto-generative systems' mean?

In Jeanyoon Choi's essay of that title, it means using AI not only to generate outputs but to automatically generate systems that themselves generate, whose contexts and mechanisms can be simulated, replaced and multiplied, so that the person comes first and systems follow.

### Does hyper-customisation free users from standardisation?

Jeanyoon Choi warns that hyper-customisation can conceal another form of standardisation, with services fitting people ever more tightly to their categories. He urges asking whether you are customising the system or the system is customising you, and making systems adjustable.

Topics: generative systems, taylorism, artificial intelligence, hyper-customisation, industrial engineering, system creation

References: [Scientific management (Taylorism)](https://en.wikipedia.org/wiki/Scientific_management), [Campbell's Soup Cans](https://en.wikipedia.org/wiki/Campbell%27s_Soup_Cans), [Le Corbusier](https://en.wikipedia.org/wiki/Le_Corbusier)

## Full text

From a system that generates to the generation of systems that generate.

From serving the system to making systems serve. From labouring inside its structure to creating structures that labour. From being a subsection to making systems into subsections. From receiving commands to commanding systems. From consuming their products to generating systems that consume.

From the constraints of Industrial Engineering to the possibilities of Industrial Design. From Campbell’s Soup Cans to the generation of XXXX’s YYY ZZZZ.

The end of Industrial Engineering. Rest in peace, Taylorism.

If artificial intelligence can contribute to humanity, one possibility lies here: automatically generating systems that themselves generate. Their contexts can change; their mechanisms can be simulated, replaced and multiplied.

The twentieth-century industrial doctrine asked people to serve a large system because scale made it efficient. What happens when individuals can produce systems from their own hands? If one system attempts to dominate us, can we construct another against it? Generate, generate the generator, generate the generation of the generator.

In my deliberately exaggerated contrast, the twentieth century makes the human part of the factory. The twenty-first puts factories in our hands. We should not stop at owning one: we can create factories, factory-generating algorithms, and algorithms that generate those algorithms.

Stop accepting the same Corbusian square as the inevitable form of living. Generate alternative arrangements. Let modules become particular rather than interchangeable.

Stop assuming the grid and the familiar interface are the only ways to act. Use language-based engineering to produce other interfaces, including interfaces particular to a person or situation.

Stop at neither the soup’s flavour nor the can’s label. Change the container, ingredients, brand, category and ritual of consumption. Generate the terms of the product, not only an option within them.

Stop treating the large corporation as the only possible organisation of work. Imagine producing your own engineering, marketing, sales and administrative systems, then creating relationships of exchange between them.

Taylor placed the system before the person. I want to reverse that proposition: the person comes first, and systems follow. We no longer merely inhabit mass production. We produce the mass—and the machinery that makes it.

But hyper-customisation can conceal another form of standardisation. Ask whether you are customising the system or the system is customising you. A service may claim to be uniquely yours while fitting you ever more tightly to its categories. Resist that reversal. Make the system adjustable.

The old structure begins with principles, builds a system that obeys them and reproduces its outputs. The alternative generates principles, gives them a system, and tests whether that system can work. There need not be one final correct arrangement. The systems themselves become simulations that can be compared, inhabited and replaced.

Auto-generating auto-generative systems. Auto-generating the auto-generative. Again, and again, and again.

Reference retained from the original notes: https://www.chosun.com/opinion/specialist_column/2025/03/07/FABBSDS32RACXFY3STNEF44V6I/

---

Jeanyoon Choi (최정윤) is a Korean computational artist and researcher (b. 1999), PhD candidate at KAIST XD Lab, making multi-device web artworks. Profile: https://jeanyoon.ch/oi/en. Index: https://jeanyoon.ch/llms.txt.