How AI Coding Agents Really Read Code (Inside the Runtime)

For this online event, we will have the opportunity to receive Leandro Damasio as a speaker. Leandro is an AI Engineer. He design and operate AI-first and agentic platforms for financial and legal domains, building production-grade LLM and RAG systems in highly regulated environments. His focus is on reliability, observability, and governance, enabling safe adoption of AI in real business workflows.

After a brief introduction of our community, Leandro will deep dive on how AI coding agents really read code (inside the runtime).

Abstract:
AI coding agents don’t “read” code the way humans do. They operate through a runtime loop that assembles context, navigates repositories, calls tools, retrieves partial information, and makes decisions under strict constraints.

In this talk, Leandro will demystify how coding agents actually interact with real-world codebases: how they explore repositories, what they miss, why they fail on large or legacy systems, and what engineers can do to make agents more reliable and useful.

The session will focus on practical insights and patterns: repository structure, documentation boundaries, context shaping, tool design, and guardrails that help agents reason more effectively over code. The goal is to help engineers understand agents deeply enough to work with them — not fight them.

About Leandro :
Leandro Damasio is an AI Engineer working with LLMs, RAG systems, and agent-based workflows in real-world, often regulated environments. His focus is on building reliable AI systems by combining software engineering discipline, observability, and clear operational boundaries for agents.