You’ve probably tried letting AI design your interfaces. You give it a prompt, it spins up a mobile login page, and the result is a visual disaster: random colors, guessed spacing, and a layout that looks like it was assembled by a blindfolded intern. We blame the model for being dumb. We think it lacks creativity.
But that’s not the issue. The problem isn’t that AI doesn’t know how to draw. The problem is that it doesn’t know what “correct” looks like.
AI doesn’t have a creativity problem; it has a blindness problem. It’s painting in the dark and hoping we like the picture.
We are terrified of AI replacing engineers and designers. We think the competitive edge in AI tooling is how many APIs you can hook up. It’s not. The real moat is a set of hardcoded disciplines that force the AI to look before it leaps, and look again after it draws.
Look at the penpot-uiux-design skill. It doesn’t just hand the AI a brush; it enforces a brutal three-step rhythm: probe before you create, ask for the design system before you drop a single pixel, and export the result to look at it when you’re done. Screw up this sequence, and the output quality collapses.
This is a massive twist on how we view AI automation. We assumed the hardest part to automate was the mechanical act of drawing. We were wrong. The hardest part to formalize is the invisible judgment that tells you when the drawing is finished.
A tool that generates without verifying isn’t an assistant; it’s a slot machine. You just pull the lever and pray for a jackpot.
The real value of this skill isn’t the generic “Golden Rules” of design—anyone can tell an AI that bigger elements mean more importance. The true value is in the five API gotchas hardcoded into the documentation. Things you only know if you’ve bled over them: width and height are read-only, flex child arrays reverse themselves in columns, and text.resize breaks if you don’t manually reset growType.
These aren’t things you can deduce from reading a syntax manual. These are scars. And by writing them into the AI’s mandatory workflow, the skill’s creators did something brilliant: they turned human pain into machine discipline.
The true value of an AI skill isn’t the code it can write; it’s the human suffering it managed to encode.
For the engineers and designers sweating over their relevance in the AI age, this is your answer. If even judgment can be encoded into a rigid process, what’s left for us? The act of discovering the next set of rules. The AI doesn’t know that a button needs a 44x44px minimum touch target, 3:1 contrast, and four interaction states. It only knows what we have explicitly told it. The human edge is knowing what “right” means before the machine ever asks.
But don’t worship this specific tool just yet. The underlying repository is already archived. You’re getting a frozen snapshot of brilliance, not a living ecosystem. The external reference files are hosted on fragile web paths that could rot tomorrow. It’s a blueprint, not a foundation.
Take the blueprint. Steal the discipline. Stop obsessing over how many tools your AI agent can call, and start obsessing over the guardrails you put around them.
Don’t fear AI taking your job. Fear the day someone successfully encodes all your hard-won, painful, tacit knowledge into a rigid, unbreakable AI workflow. Until then, your mistakes are your moat.
FAQ
Q: If the AI can draw the UI, what's left for human designers?
A: Judgment. AI doesn't know what 'correct' looks like. It only knows what you explicitly tell it. The human edge is knowing that a button needs a 44x44px touch target and 3:1 contrast without having to look it up.
Q: What's the practical takeaway for engineering teams?
A: Stop obsessing over how many APIs your AI agent can call. The real value is in encoding your team's 'gotchas'—like reversed flex arrays and read-only dimensions—into rigid, unbreakable workflows that force the AI to verify its own work.
Q: Is the Penpot UI/UX design skill ready for production?
A: No. The underlying repository is archived, meaning you're getting a frozen snapshot of brilliance, not a living ecosystem. Use it as a blueprint to build your own internal constraints, not as a foundational dependency.