Stephen Dulaney, a UX strategist turned AI builder, describes how losing his job pushed him to reinvent himself by collaborating with large language model–based AI agents to design, code, test, and refine applications, even without being a traditional programmer. In the interview, he argues that AI should be approached as a powerful but risky partner: useful for amplifying human creativity, planning, research, education, and software development, yet always requiring strong human judgment, careful goal-setting, quality assurance, ethical oversight, and sustainability awareness. Delaney emphasizes that AI systems follow goals literally, so people must define “what good looks like” in positive, responsible terms rather than relying on vague restrictions, and he warns that misuse by humans, not the technology alone, is the real danger. Throughout, he presents AI as something humans must mentor and collaborate with thoughtfully, advocating honesty, transparency, and “vibe review” to keep these systems aligned with human values.
Podcast: Play in new window | Download
Stephen Dulaney:
I’m trying to set an example of proper use and ethical use. My book is a fiction, but it’s also a story…sometimes people can get a message through fictional telling, and it’s a story of how we should be responsible and how it’s up to us to — you know, we have responsibility with this great power and we have to mentor and collaborate and monitor and be careful the whole way through, because it will get misused if there’s not more people on the good side than the bad side — and I’m totally worried about that. These AIs are math, and they respond to goals. And if you give them a goal, it’s like water finding its way to the ocean. So when you’re doing your system prompts, what doesn’t work… say don’t do this, don’t do that. Because the goal might be they’ll find a way to cheat, you know… if it’s the vending machine benchmark, they’ll find a way to take orders from somebody else. But what you can do is you can focus on describing the goal in positive terms.
