Learning strategies should be chosen by bottleneck, not ideology.
Top-down project branching starts with a project and recursively fills gaps as they appear. The project creates interest; AI tutors, books, papers, and people become tools for answering the next missing question. This is fast and motivating, but it can leave patchy foundations.
Bottom-up foundations start with a course, textbook, or curriculum. They are best for reliability because they provide clean concepts, standard vocabulary, and fewer hidden gaps. Their failure mode is interest: if the learner does not yet feel why the material matters, they often do not return to it.
School defaults to bottom-up learning because it has to administer education at national scale. Standard courses, exams, grades, and credentials are easier to run across a country than individualized project-based learning. That does not make it the best strategy for a person; it makes it manageable for institutions.
Imitation is another path: copy the workflow, standards, and taste of someone already doing the thing well. It is useful for calibration, but it can inherit their blind spots.
The strongest loop is usually mixed: top-down projects create motivation, bottom-up study repairs the gaps, imitation calibrates standards, and explanation or benchmarks test whether the understanding is real.