Short, research-informed writing on Claude, prompting, agentic systems, and the broader field. Between book editions, we keep writing.
The failure patterns that appear in almost every beginner session — and how to fix each one.
The word is everywhere and means different things depending on who's using it. Here's the precise definition we work with.
Claude Code is not a better chat interface. It's a different kind of system that requires a different way of thinking about AI assistance.
A researcher's guide to calibrating trust: what Claude does reliably, where it consistently falls short, and the verification protocols we use.
Patterns and anti-patterns from our research into multi-agent architectures. Coming soon.
An honest look at the current state of LLM hallucination research and what it means for practitioners. Coming soon.