Your AI Can Design. You Still Have to Review It.

In 2018, Angel gave a Devnexus talk called "When Developers Must Become Designers." The premise was simple: if there's no designer on your team, you need enough UX knowledge to recognize what works, what doesn't, and why.

Nine years later, it's the same problem with a new cause: AI can generate the interface for you. Describe a screen, get a polished UI, ship it. Vibe coding has made skipping the design process easier than ever.

But a finished-looking interface can hide unfinished thinking. AI can know your design system, follow your standards, and understand UX principles, and still make bad UX decisions. We already review generated code for exactly that reason. Most teams aren't reviewing the generated experience.

The failures are surprisingly predictable. AI builds the screen you described instead of the one the user needed. It handles the happy path you mentioned and misses the edge cases you didn't. It produces something coherent in isolation that falls apart when real data, real users, and real mistakes show up.

We'll give you a practical way to review AI-generated interfaces without becoming a UX designer. We'll look at the UX principles generated interfaces most reliably break, walk through a 10-minute review you can use on your next AI-built screen, and learn to separate subjective design preferences from actual usability problems.

The tools changed. The need to recognize bad UX didn't.

30 to 60 minutes. Can be delivered solo or jointly with Susa Hart. A follow-up to "When Developers Must Become Designers," delivered at Devnexus in 2018. No laptop needed, all examples are shown from the stage.

The Leadership Skills Your Day Job Won't Teach You.

You do not have to be a manager to lead. Whether you want to manage one day or stay an IC, you need to be able to run a project, align people who do not report to you, and hold a group to a decision. Nobody teaches that, and most jobs will not hand you the chance to practice it.

Community work and side projects will. That is where I learned it. Seven tech communities, where nobody can be fired, nobody is paid, and everyone can leave whenever they feel like it, so the only tools you have are the ones that do not come with a title. I took those to Microsoft, where they showed up in a product launch, a global hackathon, open source AI projects, and a tour that landed in seventy cities worldwide.

This is what each of those taught me and what it looked like when I used it at work. It matters more now, because AI is absorbing the small, low-stakes projects people used to learn on. AI can write the code. It still cannot get five teams to agree on what to build.

You will leave with:

- What running something on the side teaches that your job does not
- How to pick a project that builds the skill you are missing
- How to describe side project work in terms of the skill it built, rather than what it produced