Future of UX

There are genuinely useful tactical ideas - structured evals as rubrics, layered Claude MD file architecture, and the 'agents are just a series of files' reframe - but they are spread thin across heavy conversational padding and host affirmations. The core insights are surfaced but not developed to depth a senior operator would need to act on.

“I think in a funny way I'm cutting fewer corners because I'm actually able to get AI to fill in all that. I would have deprioritized because I would have said, oh, I need to deliver this by tomorrow, I don't have time.”

“evals are just a set of measurements. It's like a rubric.”

One genuinely fresh reframe - 'start thinking of agents really as just a series of files' - and the idea of running structured evals before live insights work are mildly counterintuitive; everything else (document your knowledge first, don't let AI decide for you, check outputs) is standard AI-workflow orthodoxy recycled in UX language.

“start thinking of agents really as just a series of files”

“If I feel like I haven't thought hard during the week, then I realize I have to cut out an AI workflow”

Caitlin is a real practitioner who has held a Head of UX Research role and built genuine Claude Code workflows, giving her credibility over pure thought-leaders; however, her experience is primarily solo/freelance at small-scale startups, she pivoted to this specialty partly by necessity, and she has no notable track record at significant organisational scale.

“I was a head of user research in a SaaS company before I kind of went freelance”

“the thing I kept getting requests for was around the stuff I was posting on LinkedIn, which was all around how I was experimenting with AI for different research use cases”

The WHOOP folder walkthrough is the episode's only concrete artefact, but the guest explicitly states all the data is fake and synthetic; there are no real client names, no measured time savings, no accuracy percentages from evals, and no before/after metrics - just named tools and described file structures.

“this is a fake folder. To be clear, I don't work with woop, I never have. So this is entirely fake data”

“run the same process again and then check if all of the numbers, all of the counts for all of the multiple choice questions, like the measure is the counts from the multiple choice questions and whether they are all the same”

The host asks useful clarifying follow-ups (how MD files reference each other, where skills are saved, what a pull request is) that help listeners unfamiliar with the tooling, but there is no substantive pushback, no challenge to any claim, and a high volume of unearned affirmation that prevents the guest from being pushed to greater depth or precision.

“I love that. I love that because last year I also had this agent rabbit hole”

“How do you Reference the different MD files to each other”