"While AI-generated images aren't always clean and perfect either, their imperfections feel different. A human error is a small mistake in execution, like a shape that's slightly off, or a line that doesn't quite close. An AI error is a hallucination, something that looks wrong in a way that's random and disorienting when you notice it."
Handmade Designs: The New Trust SignalHere's something I keep noticing in conversations with founders, product leaders, and hiring managers. There's this growing confidence that AI has solved design, that the tools have gotten good enough that you can prompt your way to a solid product, ship it, and move on to the next thing. And honestly, I get why people feel that way, because the output really is impressive, the layouts are clean, the spacing is considered, the type choices are reasonable, and everything works on mobile without anyone having to think too hard about breakpoints or grid math. But something is being lost in that confidence, and I think it matters more than most people realize.
"We believe the next trillion-dollar company will be built by fewer than ten people."
Sequoia CapitalTara Tan makes the case that the best product work happens when designing and building aren't separate phases but a single, continuous loop. The tighter that loop, the better the output. It's something most experienced practitioners know intuitively, but she articulates the mechanics well: feedback latency is the enemy of quality.
This resonates with how I think about the role of design tools that ship real code. The gap between "what I imagined" and "what got built" is where quality dies. Anything that shrinks that gap, whether it's a better handoff, a shared component library, or an AI that generates production-ready UI, is worth paying attention to.
Gary Tan, Andrej Karpathy, Francesco Bonacci writing about "Vibe Coding Paralysis." The signal is everywhere... AI isn't just changing how we work, it's changing how our brains feel after work.
A Boston Consulting Group study of 1,488 workers put a name to it: "AI brain fry," a mental fog from excessive oversight of AI tools that leads to difficulty focusing, slower decisions, and headaches. Productivity peaks at three simultaneous AI tools and then drops. Meanwhile, Axios reports that work with AI agents is "starting to look less like a fun quirk and more like a pathology."
We don't fully understand the impact AI is having on us as a cognitive system. The technology has the capability, but I'm not sure we have the cognitive capacity to keep up.
Design has always done some PM work. PM has always done some design thinking. Designers code a little, engineers roadmap a little. The roles have always had overlap, and some will argue they were always blurry. Maybe, but overlap and collapse are two different things.
Overlap means you can reach across the aisle, collapse means the aisle disappears, and this era of AI is going to stress-test which one we're actually looking at. This isn't without precedent. In graphic design, typesetters, prepress operators, and the people who set up print were all distinct roles — until software collapsed them into one person. Nobody debates whether those roles should come back. They just... went. So why would this time be any different?
Economist Noah Smith argues that AI isn't shrinking opportunity, it's reshaping it. His case rests on comparative advantage, arguing that even as AI gets better at all tasks, it's constrained by compute, so we have to choose wisely what it does and doesn't do. As a result, the demand for human judgment, coordination, and decision-making grows with it.
It's the clearest counter to the "we're all getting replaced" narrative I've seen, and I think he's largely right. Yet, the transition won't be evenly distributed. design leaders especially should be paying attention to which parts of their role are overlap and which are core.
"There are two exponentials. There's the exponential of making the model smarter, and there's the exponential of figuring out how to use the model. And I think the second exponential is actually faster than the first."
Dario Amodei on The Dwarkesh PodcastEvery few months someone declares visual editors dead. Then a new one launches, raises a round, and gets traction. The reason is simple: most people think visually. Code-first workflows optimize for precision and repeatability. Visual tools optimize for exploration and speed. These aren't competing values, they serve different moments in the same process.
The real shift isn't visual vs. code. It's that the best tools are collapsing the gap between the two. Figma ships Dev Mode. Framer ships to production. Bolt and Lovable generate full apps from prompts. The winners won't be the ones that pick a side, they'll be the ones that let you move fluidly between both.
Surprisingly, it wasn't an exploit or a prompt injection.