Evals as PRDs, Agent Observability, and Anthropic's Rough Week

Three threads worth your attention today: Ankur Goyal at Braintrust made a strong case that evals are the new PRD (the product spec for AI systems), LangChain dropped a batch of production-focused content on agent observability and multi-agent patterns, and Anthropic's week got complicated in ways that matter for anyone building on Claude.