AI Product Design - The Last Symbol of Humanity
An experiment in what happens when one designer suddenly has a lot more arms.
I came across a line from designer Elliott Earls that stuck with me: “What is a designer now, after AI? A symbol of humanity.”
I liked it, but I didn’t want to take someone else’s sentence and pretend it was mine. I also wasn’t sure I was done with it. So I threw it into AI, pulled at it, made it more dramatic than necessary, and eventually landed on The Last Symbol of Humanity. Then came the octopus.
There wasn’t a sophisticated reason for it. I wanted something strange at the centre of the experience, generated footage with Kling AI, and this creature floating in a dark ocean had the right feeling. I started building an experimental landing page around it: enormous typography, very little UI, and mostly an excuse to see where the thing would go once I could actually interact with it.
Only later did the octopus become a decent metaphor for how I work with AI now.
Octopuses have this weird distributed nervous system, with a lot happening in their arms. That felt familiar. I can have Cursor working through code while I’m generating footage somewhere else, talking through an idea with ChatGPT, investigating a compositing problem, changing typography, then jumping back because something I just saw changed the idea.
Maybe that’s one of the skills designers need now: working with several agents and different kinds of problems at once, without losing the thread in the middle.
An idea, then something I can poke, then something I understand better
I started with the title, the octopus video created with Kling AI and a feeling: deep ocean, entanglement, something slightly uncanny. Then I built it in Vite and React through Cursor.
That’s increasingly what code is for me. I’m not trying to become a software engineer. I care about it because it shortens the distance between having an idea and having something real enough to react to. And once it was real, it started misbehaving.
I’d removed the black background from the octopus footage, which also removed its black eyes, so now the giant title and gradient were visible through its eyeballs. The glass labels developed a strange colour seam. Moving the octopus made typography that looked right five minutes ago suddenly feel wrong.
None of this existed in the idea. It appeared because the thing was running.
I built myself a little design desk
At first I was making visual changes through code and Cursor, but asking an agent to move something left, then a little more, then actually halfway back gets stupid pretty quickly. Especially when it takes longer than doing the same thing in Figma. In this case, AI was somehow making a simple visual adjustment less efficient, which felt backwards.
So I added Leva and turned it into a control panel. I could move and scale the octopus, switch clips, change the gradient, typography and tracking, or position the black shapes behind its transparent eyes.
Instantly, without burning tokens on every tiny adjustment. I was basically building myself a mini Figma, except custom to this experiment and directly connected to the thing I was making. A little programmatic design environment built around the work itself.
That felt very refreshing. It still does.
That gave me a better split. AI could help me build the machinery, then I could use my own eyes again. Move it. Nope. Back. Bigger. Too much. Slightly less. Better. I wasn’t asking AI to make every design decision. I was using AI to make the design more explorable.
Following the stupid problems
The eyes needed adjustable black shapes underneath them. The green-screen footage of another video still had a fringe, which sent me into ffmpeg, chromakey and despill. A mysterious line through the glass labels eventually turned out to involve writing-mode and browser painting.
Tiny problems, but this is where the experiment became interesting.
Getting a plausible first version with AI is becoming incredibly cheap. The interesting work starts when there are letters showing through your octopus’s eyeballs and nobody knows why. You still have to notice it, investigate it, try something and decide whether it actually worked.
I also kept losing combinations I liked in Leva, so I added “Apply to code” and a way to copy the current values straight into an AI conversation. The machine could remember the numbers; I could get back to deciding whether the thing actually looked good.
At that point, I realized I wasn’t just building a landing page. I was building a workflow around how I wanted to design it. It also left me with a bunch of ideas for what a design system could become if the workflow itself was part of the system, not just the components.
But that’s for another experiment.
The octopus and the AI product designer
By the end I was moving between Cursor, Kling, React, Leva, ffmpeg, browser tools and AI conversations. Calling that “I made a landing page with AI” feels almost useless.
What changed was how far I could follow the idea myself. I could reach into code, imagery, interaction, typography, compositing and video without pretending I’d suddenly become an expert in all of them. AI gave me enough reach to investigate something, make it tangible, understand a little more, then move somewhere else.
That’s probably why the octopus stuck.
AI can give almost anyone more arms now. The harder part is knowing where to send them, spotting when they come back with nonsense, and pulling everything into something coherent without losing the reason for using AI in the first place: to ship real and better things faster.
Maybe that’s where the humanity in Earls’ line sits for me. As more of what people experience is shaped by machines, the designer becomes the one still responsible for keeping those experiences human.
Not by resisting the machines, but by staying involved: noticing what feels off, questioning what gets generated, changing direction, and fixing it directly in the work. Being able to do that in code feels like a gift.
I may be biased, but I think we’ll need more of that, not less. We’re not far from a future where a lot of what people experience is shaped by machines and could easily start to feel more disconnected.
In that sense, the AI product designer becomes one of the last symbols of humanity inside the system—not trying to imitate what’s human, but understanding the real world humans actually live in.
Comments