In 2021, before ChatGPT existed, we let an AI write an entire article from our readers' intent, to show Quantum UX working in the wild.
This article is a 2021 experiment: its body was generated by AI (an OpenAI model plus a proprietary Python engine) from the keywords, social comments and site content around Quantum UX, with the author only styling and linking it. The point was to demonstrate Quantum UX (generating an experience on the fly from user intent) years before generative AI went mainstream. The output was fast (about 1.5 minutes) and stylistically convincing, but also rambling and factually wrong in places, which is exactly the lesson: AI needs expert oversight.
Key takeaways
- The article's body was AI-generated in 2021 from user intent, not written by a human.
- It demonstrated Quantum UX: an experience generated dynamically from many people's inputs.
- It foreshadowed generative AI (ChatGPT arrived over a year later).
- The AI mimicked the author's style but also invented errors (animals driving cars, a misattributed quote).
- The takeaway: generative AI is powerful but needs expert judgment to catch what it gets wrong.
The experiment
Months after publishing on Quantum UX and XMI, the team gathered the search keywords used to find the page, the social-media discussion around it, and its own archive, and fed all of it to two AI engines. The result was a full article written in the author’s voice, produced in roughly a minute and a half. The only human work was styling, links and minor punctuation. Generating an experience from aggregated user intent is, in essence, the definition of Quantum UX.
The ideas worth keeping
- Quantum UX needs multiple actors interacting at once; it analyzes multiform (multidimensional) data, unlike one-dimension-at-a-time methods.
- XCI can be framed around different actors (for a car: Driver XCI vs Vehicle XCI), which changes how research and design apply.
- A broad definition of UXD: designing experiences among humans, and between humans and non-human 'hyper-entities' that behave like humans.
- That implies designing for perceptions, including empathy with machines and how they 'perceive' themselves and us.
The honest results
As a quick experiment, it had errors: it suggested animals driving cars and misattributed the memoir ‘Across Patagonia’ to James Norman Hall (it was written by Lady Florence Dixie). The AI understood the variables, concatenated them and faked the author’s style convincingly, but a human still had to catch the mistakes. That is the whole argument of our later essay on AI expertise.
Why it mattered
The same technique, combined with analytics and retargeting, could serve a statistically validated landing page, a culturally tailored experience, or a store showing exactly the right products, which is what Amazon and Alibaba already do. Done in 2021, it was an early, concrete glimpse of the generative-AI era. As the piece put it: AI and UX is the future, and the future is already old.
Yes. The body was generated in 2021 by AI from reader keywords, social comments and the site’s own content. The author only styled, linked and lightly corrected it.
Generating an experience on the fly from aggregated user intent is the essence of Quantum UX, so the article is itself a working example of the framework.
Yes. It suggested animals driving cars and misattributed a book. It mimicked the author’s style well but needed human oversight to catch factual errors.
A non-human entity that behaves like a human (a robot, an assistant, an AI agent). The article argues UXD should design for and empathize with these, not just human users.
It demonstrated generative, intent-driven content over a year before ChatGPT, showing both the power of the approach and the need for expert judgment.
Want AI used with expert judgment, not blind trust? Talk to Dorve.
Start a project