This blog post discusses the safety considerations involved in coding agents, particularly focusing on the roles of models, runtimes, and tools in agentic software development. It highlights the risks associated with default setups and provides insights on how developers can enhance the safety of their own systems while understanding their limitations.