After talking to designers, one of the biggest pain points of designing with agents is how much effort goes into giving them the right context.
Right now, it means manually feeding them design files, docs, and instructions every time you start something new. And starting from scratch after each task.
But your design tools should already understand your product. Not just your design system, but the research, decisions, and reasoning behind it.
Building software with agents has made me appreciate the design process so much more, not less.
A couple friends asked about my current workflow and how design fits into it, so I wrote it up.
@paper has been a game changer, giving design a proper place in an increasingly code-centric process.
atifaz.am/writing/the-softwa…
Been experimenting with training smaller models on internal product and design artifacts.
Really interesting potential around cost and performance compared to relying on frontier models for everything.
Curious if any design teams are exploring this too. Would love to compare notes.
I’ve been tinkering on design systems tooling bit and it made me think
I wonder if AI makes design systems less of a team and more of a property of the product
The system still matters
But if AI can enforce patterns and validate components, I’m not sure every org still needs a dedicated pillar around it
The past decade was pretty much a golden era for product design, but it also created a generation of designers who learned to “assemble” interfaces more than design them.
Move some components around. Connect the screens and ship the flow.
That worked when design and engineering were separate roles.
AI is clearly collapsing that boundary.
The designers who thrive from here won’t just know how to arrange software. They’re going to need to know how to build it.









