Who I am

I'm a technical leader who has been working in and around AI for more than a decade, from the earlier era of trained models built on labeled data, recommendation systems, computer vision, NLP/early chatbot experiments through today's foundation models, agents, and generative AI platforms.

Along the way, I started noticing a pattern. Every major leap forward seemed to produce its own category of failure. Funny, absurd, occasionally alarming, but usually rooted in the real limits of the current state of the art.

I've always had a soft spot for these fantastic trainwrecks.

Hallucinated case law, rogue chatbots, demos that fall apart on stage, confidently wrong outputs. They aren't just jokes. They show you where a system is brittle, where the marketing has run out ahead of the engineering, and where the next round of fixes is probably going to come from.

This site is where I keep track of them. The format is inspired by Molly White's excellent Web3 is Going Just Great, which brilliantly created the concept.

A note on tone

This isn't a site about dunking on AI. I work in the field because I think it matters.

But if we're going to put these systems into software, search, medicine, law, schools, and government, we should also be honest about how they break. The failures are part of the story. Often the most useful part.

Disclaimer

Personal project. The views here are mine. It's not affiliated with, endorsed by, or sponsored by any current or former employer, customer, or partner.

All editorial choices are mine. So are the mistakes.