Tag Archive

#evaluation

Production AI Agents: Tool Boundaries, Evaluations, and Human Oversight

A practical architecture for dependable AI agents: narrow tool contracts, explicit state, bounded execution, adversarial evaluation, and human approval for consequential actions.

Nov 4, 20268 min read

What a New Agentic Chemistry Workflow Teaches Us About AI in Production

A close engineering read of Google DeepMind’s November 2026 cheminformatics demonstration—and the practical lessons for domain tools, evidence trails, and scientific review.

Nov 4, 20267 min read