June 16, 2026
I read Mollick in the morning and Linas at lunch, and what is sitting on my desk is the same observation arriving from two angles.
Linas’s beautiful new design stack is, in Mollick’s terms, exquisitely frictionless. It is built to remove every place where a designer used to have to stop and choose. The font is chosen for you. The component is chosen for you. The dashboard layout is chosen for you. The brand voice is generated alongside the brand mark, in twelve consistent variants, before you have made a single decision about who the product is for.
This is wonderful if you have already made all those decisions in your head and you are using the tool as a hand. It is catastrophic if you have not, because the tool will make those decisions for you, plausibly, and you will not notice. The output will not look wrong. It will look like a product. It will pass every recruiter scan, every investor first glance, every internal review. And six months later, when the user does not stick, nobody will be able to point at the specific decision that was wrong, because there was never a decision in the first place. There was only a default.
I have watched this happen four times in the last month, in critiques with my students. The work that comes back from a week of Claude Design and image gen is almost always more polished than what the same student produced six months ago by hand. It is also, often, more anonymous. Three projects in a row look like they were designed by the same invisible studio, because in some sense they were. The work has been beautifully executed by a system that has no idea what the student was actually trying to say.
This is the design version of Mollick’s meaning-shaped writing. We are makinginterface-shaped products that take mental effort to navigate and give the user no equivalent understanding of what the product is for or why it exists.
The design engineer, again
I have been making the case for a while that the most interesting professional in this moment is the person who can move fluently between the model, the interaction, and the build. I want to refine that argument now in light of these two readings.
The design engineer is not interesting because they can do what Claude Design does. Claude Design will keep getting better at that, and so will whatever comes after it. They are interesting because they are the person who decides where, in a stack that wants to be entirely frictionless, friction needs to be put back in on purpose.
That is a design skill, not an engineering skill, although it is impossible to exercise without engineering literacy. You cannot decide where to put friction back into an AI product if you do not understand what the model is doing, why it is wrong sometimes, where a user’s mental model will diverge from the system’s behaviour, and what trust feels like when it accumulates and when it collapses. The Taipei version of the AI tutor is a design decision dressed as a system prompt. The Turkey version is the default.
A specific working hypothesis I am sitting with: in the era of one-shot brand identities and one-prompt prototypes, the designer’s job is less about producing the artefact and more about deliberately authoring the moments where the user, the team, or the designer themselves are forced to stop and choose. The choreography of friction. The placement of the pause.
This is the work that does not get easier when the tools get better. It might be the only design work that survives the next eighteen months unchanged.
One thing I keep wanting to ask
If you build something with Claude Design and ChatGPT Images 2.0 and ship it in an afternoon, what part of it would you be able to defend in six months, when a user is angry, when an investor is skeptical, when a teammate inherits the file? Which decisions in the artefact are yours, and which ones did you accept without noticing?
I do not have a clean answer, and I am suspicious of anyone who claims to. But the question feels like a useful one to carry into the rest of the year.
What are you choosing to keep human in your own design work? Reply and tell me. I am collecting answers.
Sources
Ethan Mollick, Choosing to Stay Human, One Useful Thing, 26 May 2026.
Ethan Mollick, Sign of the future: GPT-5.5, One Useful Thing, 23 April 2026.
Linas Beliūnas, Anthropic Just Told AI Founders Exactly What to Build in 2026, Linas’s Newsletter, 4 May 2026.
Linas Beliūnas, Turn ChatGPT Images 2.0 & Claude Design Into Your Chief Designer, Linas’s Newsletter, 24 April 2026.