Long-form essays on building agentic AI systems you can trust: where they break, and how to engineer around it.https://tomsullivan.dev/en-us