Does reconstructing code count as active recall?

I’m building AlgoBite around a deliberately guided interaction: you see all the Python chunks, but still have to rebuild the control flow and pass the tests.

I can see two honest readings:

- it forces retrieval of structure, state updates, and guard clauses

- visible chunks make it recognition with extra steps

I do not think it replaces solving from scratch. The narrower question is whether it creates enough retrieval to keep patterns fresh between interview cycles.

Where would you draw the line between useful retrieval and glorified recognition?