At Rails World 2026, DHH said that writing code by hand is no longer economically viable. That the days of hand-coding are over.
DHH, creator of Ruby on Rails, at Rails World 2026
The point is that while we were debating the risks—and, yes, the moral implications—of handing code generation over to AI, “most people at most companies” were already adopting these tools.
And they were doing it at a remarkable pace.
We can speak to what we’ve seen ourselves: using AI to design and develop digital products consistently delivers results beyond our expectations, in time frames that would have been hard to imagine until recently.
Here’s an example. Recently, while exploring UX approaches for software that controls machinery, we built interactive prototypes directly in HTML. We needed people to use the interface as if it were real, to get a feel for its usability and ergonomics. Until recently, bringing a developer in at that stage would have been prohibitive in terms of time and budget.
The change is just as clear in development. Once you find the right workflow, an agentic approach makes it much faster to build fairly complex projects. You can maintain a high standard of UI quality and move past many of the usual constraints.
And, to the delight of any product marketer, there’s another big shift: with or without a CMS, we can now translate, revise, and improve copy and pages in no time. Fixing SEO or accessibility issues has become faster too.
Looking beyond our studio, we see an equally profound change. Our tech clients are becoming more self-sufficient: they can start with a design system we’ve built together, develop new parts of the product, and extend the design with AI agents following the rules and patterns we defined.
Sounds great, doesn’t it?
Exciting and unsettling at the same time
AI makes two seemingly opposing things true at once.
This change is exciting because of the gains in productivity, quality, and efficiency. But it is also unsettling for people who find that part of the work they built their profession around is no longer needed—or simply no longer valued in the same way.
Beyond the self-promotional posts on LinkedIn and the people who directly benefit from AI adoption at work, I see very different reactions. There’s the programmer who found meaning in writing code. The photographer or illustrator wondering whether this will be their last project of its kind.
Of course things change. But finding out that what you did—or at least how you did it—is no longer needed can shake much more than your income.
And I don’t think there’s a realistic way to stop this change. It will keep moving because it’s too fast, too cheap, and too effective.
That said, telling someone they need to reinvent themselves doesn’t make it any easier. I think we should bear that in mind when we talk publicly about our enthusiasm for these technologies.
Attention and effort
In his article “Does a design studio still make sense?”, Matteo writes that as building things gets easier, judgment and taste become more valuable: the ability to choose, to recognize a good solution, and to understand why it’s good. Not just one that works.
But I think judgment and taste need something else: attention. They matter only if we genuinely care about what we’re doing.
We can ask an agent to fix a bug, build a feature, or explore an idea. It will probably produce something. But we still have to understand what we asked it to do, follow the choices it makes, and check what it delivers.
Delegating the execution doesn’t mean delegating our attention. Nor does it spare us the effort of doing the work. And that’s a good thing.
Simone Weil describes attention as keeping our minds open to what is in front of us, without rushing straight to an answer.
That feels especially relevant now. New technologies can take many repetitive, mechanical tasks off our hands. But they aren’t a shortcut for work that calls for creative input.
Perhaps that’s why long sessions working with AI leave us drained—maybe exhilarated by the results, but drained all the same.
And so far, that’s fine. What worries me is the temptation to let AI work in the background while we do something else. Or to have ten or twenty agents working on different things for us at once.
For some mechanical, repetitive tasks with no real added value, that can work well. But when the choices matter, we risk losing the thread and ending up with a result we can no longer fully evaluate.
That’s also why working with Claude Code or ChatGPT can hold our attention for hours. As the AI works, we keep forming hypotheses, weighing its choices, and asking whether we’re heading in the right direction.
Attention, judgment, and taste work together. They let us use AI’s speed to go deeper and improve the work, instead of stopping at the first answer that seems convincing.
Attention as an antidote to slop
Without that attention, we risk getting stuck in the endless production of AI slop: products and content generated at scale that may be technically sound but lack care and relevance. In the end, they’re just mediocre.
In design, we risk making interfaces that all look alike. Worse, we risk passing up the original choices that could make a product distinctive and give it a competitive edge. In development, the risk is different but related. We can accept technical decisions without really understanding their consequences for the product or the business. The result may be technically flawless but solve the wrong problem.
The same is true in marketing. AI can help us think through a strategy, explore alternatives, and test ideas. But a proposal that seems to work for everyone may be right for no one. We still need to understand what it means for that particular company, its customers, its market, and its constraints.
There are two other risks. The first is letting AI indulge us. We start with an idea and keep adding, exploring, producing. AI follows along, but it doesn’t stop us to say we’re going too far—that we don’t have the time, resources, or capacity to manage what we’re imagining. The second is trusting its answers on subjects we know little about, to the point where we think we no longer need experts. I’ve already heard someone say they know how much a project costs because they asked Claude.
Attention and relationships
Then there are our relationships. Sometimes we risk paying more attention to a conversation with an agent than to one with a person.
Here’s an example. A prospective client at an AI startup, juggling a hundred things at once, tells me they instinctively thank the AI when they talk to it. Then they disappear and stop replying to our messages. So much for saying thanks.
It’s just one example, but I could give plenty more. I mention it because, amid this proliferation of Claude, ChatGPT, Grok, and agents of every kind, I think we risk losing some sense of the depth of our relationships with other people.
Frankly, becoming a company of one and orchestrating an organization of chatbots isn’t what Matteo and I want. For us, relationships matter when things are going well, but especially when there are problems to work through.
Paying attention means giving someone your time, listening, and taking care of the relationship when things are easy and when they aren’t.
It also means sometimes resisting the temptation to have an agent write everything for us.
Sometimes it’s better to get a coffee, arrange lunch, meet in person, and look each other in the eye.
That helps us see whether we’re really looking at the same problem from the same perspective, or whether we’ve simply misunderstood each other through a string of brief messages.
For Moze, the opportunity I see is to use AI to spend more time on problems worth solving and on the people we solve them with. Judgment and taste still matter. But without attention, they aren’t enough.