In the Laravel and AI bootcamp I taught in Agadir, a coding agent was part of the work. But speed of production could not be allowed to hide a lack of understanding.
The principle is simple: code that a learner cannot explain is not enough to demonstrate a skill.
Look beyond the demonstration
An application can work without its author understanding every decision. A demonstration shows the result; a technical discussion helps reveal the reasoning.
Explaining what a function does, finding where a behaviour is defined and changing a rule make that understanding visible. The aim is not to catch learners out, but to distinguish what they have understood from what still needs work.
The agent as a development tool
AI can help produce, explore or fix code. It does not remove the need to describe the requirement and review the result. That responsibility remains with the person building the project.
Training needs to teach precisely that responsibility: how to ask, check and explain.
Connect practice with assessment
This is where my coaching and jury work meet. I look at the ability to understand and justify decisions as well as the result presented.
The Agadir bootcamp and its resources provide the context for this practice.