quiz · 4 min Foundations Check What is the agent loop?A Python for-loop with no modelPerceive, then decide, then act, repeated each stepA cache of previous LLM responsesThe transformer attention mechanismWhy develop against a deterministic mock LLM first?Reproducible, free runs with no API keys or flakinessMocks answer more accurately than real modelsIt removes the need for an agent loopMocks can call external tools automaticallyIn the seam policy(observation) -> action, what does the policy do?Persists progress to local storageRenders the lesson pageChooses the next action from the observationDownloads the Python runtimeWhat are the four separable parts of an agent?Tokens, weights, prompts, embeddingsPerception, policy, action, memoryInput, output, cache, logPlan, tool, retry, scoreWhen you replace the mock with a real model, what stays the same?The agent loop and the policy interfaceNothing; the loop is rewrittenThe provider API keyOnly the print formatSubmit Mark complete ← Anatomy of an Agent Controlling the Policy →