Theory-first or projects-first is one of the oldest debates in learning to build AI, and for most beginners the answer is clear: start hands-on and pull in theory as projects demand it. Doing-first keeps motivation high and makes the theory stick. But there is one honest caveat.
Why hands-on wins for most people
- Motivation - early wins keep you going; pure theory drains it
- Retention - concepts learned to solve a real problem stick
- Relevance - you learn what you actually need, not what a syllabus front-loads
This is exactly why Fast.ai built its whole approach around building first. Start a small project, hit a gap, learn just enough theory to cross it, continue.
The caveat: do not skip theory forever
"Hands-on first" is not "hands-on only". If you never circle back to the underlying ideas, you plateau - copying patterns you do not understand. The maths and fundamentals matter; the point is to learn them in context, not before you start.
A practical rhythm
- Pick a small, real project
- Build until you hit something you do not understand
- Learn exactly that concept
- Apply it, then move on
Repeat, and theory accumulates naturally around real work - the basis of a strong portfolio.
If you are applying, not building
For applied AI, this is even more true - you learn by using tools on real tasks. Strategic Application of AI in Business is built around practical application rather than theory for its own sake.
The bottom line
Start hands-on and pull theory in just-in-time - it keeps you motivated and makes concepts stick - but do circle back so you understand what you are building. Doing-first, not doing-only. Next: how to build an AI portfolio as a beginner.