AI Regulation, Open Source Reality Checks, and Agent Cost Control

Three threads worth sitting with today: Dario Amodei publicly shifts from transparency advocacy to binding regulation (a meaningful signal for product risk planning), Clement Delangue calls out the gap between open-source rhetoric and actual contribution, and LangChain publishes a practical guide for controlling runaway coding agent costs — the first concrete cost-governance framework I've seen for multi-agent tooling.