Start with one concept.

Names, assignment, and how values persist from one statement to the next.

Generated code is full of names. If students cannot say what a name currently holds, they cannot judge whether a suggestion is safe to accept.

Open Variables

Decisions: Boolean conditions, branches, and boundary values.

AI suggestions often look locally tidy while encoding the wrong boundary (for example > instead of >=). Conditionals are where that error hides.

Open Conditionals

Repeated action, accumulating state, and stopping conditions.

Novices and models both lose track of what is true after each iteration. Tracing a tiny loop is cheaper than debugging a generated one.

Open Loops

Parameters, return values, and the difference between printing and returning.

Generated helpers often print when the caller needed a value, or ignore an edge case in the contract. Students need to read the function as a specification, not as a block of syntax.

Open Functions

Ordered collections, indexing, and aggregation over items.

List processing is a common AI default. Students should be able to say whether a fragment is searching, counting, filtering, or reducing — and whether the empty list is handled.

Open Lists

Key–value lookup and updating associated data.

Models frequently confuse membership of keys with membership of values. That is an evaluation opportunity, not a reason to skip the concept.

Open Dictionaries