Every designer's profile says "ai-native" now. Most portfolios still show only the finished screens, which is exactly the part an agent can't vouch for. If the claim is about how you work, the evidence has to be about how you work.
That's why this site has two kinds of writing. Case studies are few and deep: what the problem was, what I decided, and a section in each one called "ai in the loop" that splits the work honestly between the agent and me. Notes are short and frequent: experiments, workflows, prompts that worked and prompts that didn't.
The split matters because the field moves fast. A case study about a workflow from last spring is already history. A note from last week is current practice. Together they say two different things: here is the quality I hold myself to, and here is how I'm working right now.
What stays human hasn't changed much. Taste. Trade-offs, like a feature that's correct but makes the product louder. The small interaction details that only show up on a real device, in a real hand. An agent is very good at the translation between those decisions: canvas to code, tokens to CSS, a spec to a scaffold.
The risk is sameness. If everyone asks the same models for the same layouts, everything starts to look like everything else. My guard against that is boring: real content from day one, decisions written down, and every diff read before it ships.