Noah van Lienden

Most AI products today are a chatbox bolted onto a model. The work I want to do is the step after: interfaces and experiences that build themselves around the person using them. Generative UI, personalization, voice modeling.

I care about UX more than is probably reasonable. A model can be brilliant and the thing wrapped around it still miserable to use. That gap, between the model doing the work and a person actually getting something usable, is what I care about.

For most of software’s history you shipped one interface to a million people, because that’s all you could do. Once a model can compose an experience in real time for one specific person, that constraint is gone. I feel like most products haven’t caught up to this yet.

That’s most of what Ditto is for me: pulling someone’s writing voice out of what they’ve already written, then steering rewrites that actually sound like them. Sounds simple. It isn’t. LLMs are trained on language and still weirdly bad at sounding human. Omen was the same instinct in eCommerce, the page rearranging itself around whoever was on it.

Products shouldn’t be built around one person, either. The ones that create real value are collaborative, where the interface holds the context a group shares instead of making everyone rebuild it in their own head.

So why does every AI product collapse into a chat window? Chat is great for input. It’s how people naturally say what they want. It’s worse for output. Nobody wants their dashboard as a conversation. Dense information wants lists, charts, layout. The future isn’t a chat layer over everything. It’s interfaces that follow what I’m doing in the moment, with the model doing the real work underneath, that are a genuine delight to use.