I’m the co-founder and Chief Growth Officer of QWNTL Labs, a research lab studying why AI agents break over time, and how to measure and prevent it.

I write about the reliability shift: the industry answered “can a model do this?” and is now stuck on “can an agent be trusted to do this repeatedly, over a long job, without a human watching every step?” That’s a market and economics story more than a mechanism story, and it’s the one I cover. Field reports from teams running agents in production, the economics of the gap, and the occasional note on building a lab in public.

The mechanism and the research are my co-founder Lindell Cumes’s lane. When a question is about how, I’ll point you to him.