Learn strategies for working with AI coding agents more effectively to improve outcomes and reliability for research.
Get more reliable, research-ready results from AI coding agents like Claude Code, OpenAI Codex, or GitHub Copilot. This hands-on workshop is for researchers who want to move beyond “vibe coding” and learn the workflow strategies that help agents produce reliable code and reproducible outputs for their research. We’ll practice techniques that shape an agent’s work from the start: crafting effective prompts, writing clear specifications up front (with AI’s help), capturing project conventions in a persistent context file, and defining how the correctness and quality of the output can be evaluated. Along the way, you’ll learn to interrogate the assumptions an agent makes, ask it to show evidence that its code works, correct it efficiently, and find the right balance between hands-on oversight and letting the agent plan and work independently. This workshop assumes some prior exposure to agents; basic AI agent setup and safety are covered in a separate workshop.
Prerequisites: Basic familiarity with coding, and some prior experience using an AI coding agent. Participants must have their own access to an AI coding agent (such as Claude Code or OpenAI Codex), which may require a paid subscription.
Leticia Vega
Email