Somebody still has to care
There is a lot of talk lately about taste — how, as LLMs get better and better, we are supposed to differentiate ourselves from the machines. If Claude can one-shot a beautiful website in fifteen seconds, what is left for us to do?
But you and I both already know there is massive variance in what comes out depending on who is doing the prompting. The intangible ability to encode what you want — rather than letting the LLM guess — is what separates someone who has spent years building from someone who has not. Craft and taste come with reps, not with reading about them.
So now that building software has become cheap, we try to encode taste itself into markdown files our agents can consume. There have been big leaps in that direction — tools that turn a designer's sense of quality into shareable skills, interface kits that make beautiful UIs accessible without years of training. Anyone with a subscription can now ship something that looks decent.
This leads to an inevitable question: if someone with no software experience uses every tool available to its full extent, can they build something truly great?
I have been thinking about this a lot. Hard not to when every week brings another leap that makes the previous one look slow. But I keep arriving at the same conclusion: there has to be something beyond taste. An additional filter that separates great applications from merely competent ones.
AI can write your code. It can help you move fast. It can make impossible things feel possible. But it cannot care. It cannot make your app memorable. It cannot make your app feel like it has a soul.
Some of the best apps I have used this year are structurally simple. The code was straightforward, the architecture was not clever. But they had soul. They felt like their creators cared about what they were building. Every micro-interaction, every edge case, every moment of delight was placed there on purpose — not because a model generated it, but because a human decided it mattered.
That is the thing no amount of prompting can replace. Sure, AI has gotten so good that it preempts some of these insights. It can suggest accessibility improvements before you ask. It can spot maintainability issues and flag design inconsistencies. But it does not have the capacity to care about your application. It does not know what good feels like. It only knows what good has looked like before.
You have to be the one telling it to make things better. Even if Claude can rewrite a whole feature in a split second, you still have to be the one directing that attention.
So yes, LLMs can help anyone build faster and better than ever before. But you cannot make something good unless you care. You still have to be the ultimate filter for your product's quality. You still have to put yourself in your users' shoes and understand what works for them. You still have to pick every nit, obsess over every detail, and fix every issue until it feels right.
You still have to be the one who cares.
CraftJun 8, 20263 min read
About the author
mosh
mosh is a product designer and design engineer working with design systems, LLM-powered prototypes, agent-safe interfaces, production UI, and automated workflows.
Keep reading
- The last 20% is the whole point
AI nails the first 80% of a product in minutes. The last 20% — hierarchy, edge cases, copy, point of view — is where taste lives, and it is still your job.
- AI Design Engineering: How Product Teams Ship Faster Without Losing Quality
A practical guide to LLM workflows, agent-safe design systems, production UI, and human quality gates.
- Why I left my job to start my own company
Why I left a comfortable job to build independently — and what I gained and lost.