The great flattening
Generative AI promises speed. What once took 10 steps now takes a single prompt. A journey becomes
…a jump.
But in all this collapsing of complexity, something is lost: texture. The pauses. The progressions. The little moments that made services feel considered, human, alive. We’re in danger of building things that are:
Helpful but hollow. Immediate but not intimate.
Flattened by speed. Stripped of ceremony. Delivered with “perfect” efficiency and zero resonance. Think of the Apple product box. The way it opens, the way it resists. That tiny friction says this matters.
Or think of Miyazaki’s reaction to AI “slop.” When fans generated images in a Ghibli style, Miyazaki called it “an insult to life itself.” Rolling Stone reported his objection and it underscores how much of what makes art and design meaningful is the layered conversation, craft, iteration, and collaboration. The long, collective journey behind the final image. Skip all that, and the result feels flat.
This is where the slope gets slippery. The more we normalise instant answers, the less we value the slow, textured processes that make experiences meaningful.
AI, in its rush to remove friction, risks removing meaning too.
Liminal space needs limits
The best experiences have edges. A film has credits. A painting has a frame. A journey has a beginning and end.
AI doesn’t like edges. It jumps. It blurs. Input in, output out, floating in a kind of ambient utility.
As John Maeda puts it, design has shifted “from obstacle course to teleportation.” Autodesigners on Autopilot
But teleportation removes the experience of movement; the friction, the reflection, the detours that matter.
Framelessness isn’t freedom. It’s disorientation.
And here’s the rub: design isn’t just about the end point anyway. It’s about the conversation that gets you there. The back-and-forth between a team as they make choices. The dialogue between a service and its users. The shared sense of where are we heading? That conversation is a frame too. It gives shape, context, direction.
AI often strips that out by arriving before you’ve even decided where you’re going. The jump replaces the journey and erases the dialogue that made the journey meaningful in the first place.
If we want AI-mediated experiences to mean something, we have to design for structure, thresholds and narrative arcs, even in micro-interactions. Memory and reflection that give a sense of place and progression. Without those, even the most intelligent service becomes unplaceable.
A void with a voice.
Fake touch, real ick
Worse than losing feeling is faking it. Too many AI services lean into “human-like” chatty tones, cartoon avatars, synthetic empathy.
But a wet handshake is worse than no handshake at all. I don’t want AI that feels more human. I want AI that feels more useful. More legible. More honest.
That’s the real opportunity: not to imitate us, but to support us. To give AI shape and texture so it feels trustworthy without pretending to be our friend. And to remember that the conversation is part of the work. Between teams in the room. Between services and their users. Between audiences and the things they love.
If AI skips straight to the answer, it robs us of that dialogue. And that’s the slope we should fear most: a slide into hollow efficiency, where nothing has edges, pauses, or meaning left.
However, if we design with care, with edges, with pauses, with space for conversation, then AI doesn’t have to flatten. It can deepen.