In this episode of Phoenix Cast, hosts John and Kyle dive into two cautionary tales from Kyle’s AI-powered workflow — one where he spent $70 proving that AI detection tools are fundamentally broken, and another where he nearly lost his entire CRM database to a vibe-coded update gone wrong. Kyle walks through his process of writing a Marine Corps Gazette article using AI as a drafting assistant, only to have two leading detection tools flag it as “100% AI” three times in a row — sparking a broader debate about whether “did AI write this?” is even the right question to ask. Then things get real when Kyle discovers his customer database has been wiped by a bad code push, blindly trusts AI’s hallucinated diagnosis, and burns three and a half hours chasing ghosts before realizing the data was there all along. It’s a masterclass in why human-in-the-loop verification matters, why the basics like backups still apply in the age of vibe coding, and why tokens are cheap but trust is expensive.
Links:
AI Snake Oil: What Artificial Intelligence Can Do, What It Can’t, and How to Tell the Difference
Vibe Coding: Building Production-Grade Software with GenAI, Chat, Agents, and Beyond
https://a.co/d/0j6Uj0K5