Essays on design practice in the age of generative AI
About the series
These essays began with a question that circulates in studios and lecture theatres: now that machines can produce graphic outputs, product renders, interface screens, and working prototypes in seconds, what is left for designers? The series answers by examining what design actually is when it is examined closely. It finds a practice whose most consequential work happens upstream of any artefact, in the constituting of problems; a process that is irreducibly argumentative, conducted through a web of issues that must be opened, contested, and closed by judgment rather than by criterion; an intervention into a future the intervention itself remakes; and a profession defined, in the end, not by what it makes but by what it is answerable for. The machines take a side in a quarrel the field thought it had abandoned. These essays take the other side, and explain why the consequences live there. They draw on Horst Rittel, Herbert Simon, Gabriela Goldschmidt, Vinod Goel, and Hans-Georg Gadamer, and they are written for practitioners, students, and anyone deciding what design education and design teams should now protect.
The essays
I. The Brief Always Arrives Wrong On why the profession will not run out of work simply because machines have learned to make things. The opening essay dismantles the picture of design as a fixed inventory of artefacts waiting to be produced, and locates the profession’s real work upstream of any deliverable, in establishing what should be made, for whom, and why.
II. The Web of Issues On design as argumentation, and why the process is the work. Drawing on Horst Rittel, this essay argues that designing is the traversal of a web of issues, positions, and arguments, and that closing an issue is never a calculation but a judgment, political in nature, for which someone must answer.
III. The Quarrel That AI Reopened On why a fifty-year-old argument in design theory has suddenly become a practical question. The founding dispute between Herbert Simon and Rittel, abandoned rather than resolved, returns with force: generative AI is Simon’s account of design built and shipping, which makes Rittel’s objections newly urgent.
IV. Two Stopping Rules On how designers decide that the designing is done. Simon said stop at good enough; Rittel said there is no rule, only stopping you can defend; and the tools have quietly installed a third rule that neither proposed: stop when the output looks finished.
V. The Resolution of the Argument On sketches, prompts, and the level of abstraction at which design arguments live. Through Gabriela Goldschmidt and Vinod Goel, this essay shows that the web of issues is not medium-neutral: ambiguous representations keep issues open, and a practice of text in and renders out conducts its arguments in a channel that can only close them.
VI. A Sketch Is a Question On keeping the issues open when the artefact behaves. A direct sequel to The Resolution of the Argument, carried into interaction design: a field whose subject matter is behaviour, whose behaviour never had a native sketch, and whose prototyping repertoire, from the paper prototype to the Wizard of Oz, was a fifty-year project to manufacture the ambiguous middle that generative tools now collapse. A sketch asks, a prototype answers, and a generated artefact is an answer arriving before the question.
Declaration of AI Use in the Writing Process
There is, interestingly, a close parallel between how I write and the ideas these essays explore, it is useful to describe that process. Before the AI tools, testing an idea was expensive. I would spend weeks reading, sketching arguments, and searching for flaws. Often the idea proved to be a dead end or, worse, something someone else had already published. That investment created sunk-cost bias. I chose safer questions and clung to failing ideas because abandoning them meant writing off weeks of work.
With LLMs, exploration is cheap, and so is abandonment. The ideas I pursued became better because the ones I discarded cost almost nothing. My ability to discover the question worth asking improved. That is precisely the capacity for problem constitution that ‘The Brief Always Arrives Wrong’ argued lies upstream of everything else.
The rest of the series maps onto writing just as neatly. Like design, writing divides into two fundamentally different kinds of work.
The first is criterial. Reduce the essay to the required word limit. Check spelling, grammar, punctuation, and citations. Maintain a consistent voice and style across the series. Ensure that each essay advances the central thesis. Every one of these tasks has an explicit criterion. A word limit is either met or exceeded. A citation either conforms or it does not. Because these tasks admit of a test, they can be delegated. This is writing’s layout-that-validates. For a long time, this labour obscured where the real work lay. A polished manuscript was once evidence of an author’s sustained attention. It no longer is.
The second kind of work admits of no such test. Is the interpretation convincing? Is an objection fatal or merely inconvenient? Is the argument genuinely complete, or does the prose merely give that impression? That final question is ‘Two Stopping Rules’ translated into writing. And ‘No Ultimate Test’ applies in full: no checklist can certify an interpretation or validate a claim. The only test is whether the author can defend it in peer review, in a seminar, or in response to a reader. That responsibility cannot be delegated. The claims have one owner.
This series therefore illustrates its own argument. I use Claude to edit the essays, maintain consistency of voice and style across the series, and check spelling, grammar, punctuation, and citation accuracy. These criterial tasks were delegated. Every interpretation, every judgement, and every substantive claim remains mine.