Why your codebase lies to AI?

AI coding tools promise to understand your codebase and accelerate software delivery. Yet developers keep hitting the same wall: hallucinations, context overload, fragile fixes, and AI systems that struggle to understand real-world projects.

The problem is not the model.

The problem is the assumption that your codebase contains the truth.

In reality, code is only a projection of business intent. The reasons behind architectural decisions, production constraints, customer feedback, operational risks, and business objectives are rarely visible in the code itself.

In this session, Zaak Chalal challenges one of the most widely accepted assumptions in AI-assisted development: that indexing and vectorizing a codebase is enough for AI to understand a project.

Through real-world examples and live demonstrations, he explores why AI fails on legacy systems, why context engineering is becoming the new bottleneck, and how organizations can move from code-centric AI to intent-centric AI.

If we want AI to become truly effective in software engineering, we need to stop asking it to guess the truth from code and start giving it the context it actually needs.