Steve Jobs once talked about how:
“When you’re a carpenter making a beautiful chest of drawers, you’re not going to use a piece of plywood on the back… You’ll know it’s there. For you to sleep well at night, the aesthetic, the quality, has to be carried all the way through.”
As a UX designer (and not furniture designer) the thing that has kept me up at night is hearing the “AI will replace X” narrative. Thinking about Job’s quote and this current topic, I can’t help but think the plywood test still applies. If anything, it matters more now.
IKEA didn’t replace all furniture in the world.(Though if you were a university student, it probably felt like it did. It certainly did for me.) What it did was make furniture more accessible. More affordable. More possible. And in doing so, it accidentally made craft more visible. After your third identical flat-pack chest of drawers, something shifts. You start noticing the wobble. The quality. The shortcuts. You start asking: Is there something better than this? You start asking different questions. You start to care.
AI is creating a similar moment in tech.
It’s making building faster, cheaper, more available. It’s removing friction from execution in ways that would have felt impossible a few years ago.
But for anyone who’s actually shipped products, you know building was never the hard part. The hard part was taste. Judgment. Knowing what not to build. Knowing when something is technically correct but your users couldn’t care less.
What I see many organizations getting wrong is treating AI as the thing that defines quality, rather than the thing that exposes it.
Yes, speed improves. Yes, output increases. And what gets produced isn’t always outright slop - it’s something more dangerous: slop-adjacent. Polished enough to ship. Shallow enough to forget.
When everyone can build fast, speed stops being a differentiator. What becomes visible instead is quality - or the lack of it.
You can feel it. Users can feel it. Even if they can’t articulate why.
Here’s the temptation: we’ve killed plenty of product organizations with bloated review processes and broken cultures. Wrong incentives. Endless alignment meetings. Death by committee. So when AI offers a way to ship something - anything - faster than fixing those underlying problems, it feels like success.
But it’s not. It’s a painkiller masking the symptoms while the disease spreads. You’re still building on a broken foundation. You’ve just sped up how quickly you can produce things nobody asked for.
Making things people actually use, trust, and come back to - that’s still hard. Making things that hold up when the novelty wears off - that’s still hard. Making things that still feel intentional when no one is watching - that’s still hard.
AI doesn’t remove the need for craft. It makes the absence of it impossible to hide. When the whole world can build, the real question is simple: Who’s willing to carry quality all the way through, and who’s still using plywood?
Some Recent Links (Human Curated) ❤️
Human in the Loop Framework - This is interesting framework for thinking about what “human-in-the-loop” actually means in practice. Instead of treating human involvement as a vague term, it offers a framework that discusses where judgment, oversight, and intervention genuinely matter. Worth a read if you’re building or evaluating Agentic AI experiences.
Burgers To AI - If you’re interested in sustainability and AI, I read this substack article that analyses xAI’s data center water consumption footprint. The article breaks down the comparison in relation to In-N-Out Burger restaurants and shows how the data center is-roughly (only?) 2.5× the water footprint of a single average restaurant. Something to keep in mind next time you’re craving a burger!
AI Agent Adoption - I came across a Harvard (with Perplexity) study analyzing AI agent interactions. It reveals that 57% of usage is for cognitive tasks. Knowledge workers in tech, finance, and similar fields show the heaviest engagement on complex synthesis and decision support. I’m curious to see how this usage will change in other populations. Read the full paper here.
Designing Audio and Voice Experiences - I came across this thread from Hayden Bleasel sharing Vercel’s new release of Voice AI Elements. Six ready-to-use React components for your next project. Perfect if you’re designing voice agents or audio type experiences.
SVG Pattern Builder - Tired of hunting for the perfect repeating or animated background? Try svg.designcode.io. Built mostly with Claude AI, it lets you customize shapes, colors, and animations, then export as SVG, PNG, GIF, video, or raw code — ready for Figma, Framer, Webflow, or anywhere else.
Cursor Inspiration - I came across this inspiring thread showcasing cool projects people have recently built using Cursor. My favorite example is the currency converter that somebody built in just 30 minutes using Cursor and Claude - DESPITE the exact same core functionality already existing inside Apple’s Calculator. It shows how AI now makes it trivial to scratch tiny personal itches through personalised software.
Finally - If you’ve ever felt overwhelmed by all the terms flying around — rules, skills, MCP servers, modes, hooks,… yeah, it’s a lot - this short breakdown from the team at Cursor cuts through the noise perfectly: