A July 18 post by Peter Steinberger on X, asking "Are we still talking loops or did we shift to graphs yet?", reignited a debate over how to build AI agents, exposing a fight over how much structure is needed before an AI agent is safe to trust. DeepLearning.AI's free Agentic Knowledge Graph Construction course, taught by Andreas Kollegger of Neo4j and running 3 hours and 8 minutes, has contributed to the subject circulating again, with lessons on using Google's Agent Development Kit to build a multi-agent system. However, the article notes that a knowledge graph represents relationships in data, while a workflow graph represents relationships between tasks, and they are not the same thing. Anthropic's December 2024 essay "Building Effective Agents" stated that the most successful implementations used simple, composable patterns, and its Claude documentation describes an agentic tool loop where the model asks to use a tool, the application runs it, and the process repeats until the model stops. Anthropic's 2025 engineering write-up described an orchestrator-worker setup where a lead agent plans the search, spins up three to five subagents in parallel, and then combines their findings, noting that parallelization cut research time by up to 90% on complex queries.