AI Letters #05 - Build Agents, Not Chatbots: The Loop That Changes Everything
· 3 min read
A team spent three months building an AI assistant for internal engineering docs. Clean RAG pipeline, good answers. Then someone asked it to find the root cause of a latency spike, check the runbooks, and draft a remediation plan. It answered half from training data. Made up the rest. Confidently.
The problem wasn't the model. It was built to answer, not to act.
This issue covers the architecture that changes that: the ReAct loop, how tool calling actually works at the API level, a raw Python agent from scratch, and the LangGraph version with state persistence. Part 1 of 5 in the Build Agents, Not Chatbots series.
This letter was first published on AI Letters →
Continue reading on Substack →
Interactive Data Stories
Interactive Chart
60 Years of AI Agents: The Timeline →
From ELIZA (1966) to MCP (2025) - click each event to see what changed and why it mattered for the engineering trajectory.
Interactive Chart
The 4 Agent Architecture Paradigms →
Chatbot vs ReAct agent vs stateful agent vs multi-agent - click each paradigm to see the loop, code pattern, and when to use it.
Interactive Chart
Agent Performance Evidence Dashboard →
SWE-Bench trajectory, model comparison, and the cost paradox - the data behind the agentic AI transition, fully interactive.