New Research: How AIs Imagine Software Engineering

Phanvic's first research project examines how AI systems imagine software engineering work, from problem decomposition to implementation strategy.

The study looks at how models form internal plans, identify constraints, describe architecture, and recover from ambiguity. These patterns matter because AI coding systems are becoming collaborators inside real development workflows.

By understanding the way AI systems represent software tasks, we can design better interfaces, evaluation loops, and safety practices for agentic engineering systems.