Succeeding with Agentic Coding

Ever handed an AI agent a task and come back to a mess? Agentic coding tools are powerful. They are also easy to misuse. Without structure, an agent will happily wander off. It rewrites code you didn’t ask it to touch, then declares victory on code that’s broken.

In this session, Michael Kennedy walks through how he uses Claude Code to ship real software. You’ll see how he sets up a project so the agent has the context it needs, and how he scopes work into pieces an agent can actually finish. He’ll also cover the guardrails that keep it on track. The agent plans before it writes code and checks its own work. You learn to spot when it’s drifting, so you can stop it and take the wheel.