From zero, learning AI properly takes most people well over a year: roughly 3–4 months for programming and data basics, 3–4 months for machine learning, 2–3 months for deep learning, and several more months for transformers, LLMs and deployment. Using AI tools takes weeks; understanding and building AI systems takes longer, and the pace depends far more on consistency than on talent.
How long does each stage take?
| Stage | Typical time from zero* | Milestone |
|---|---|---|
| Programming and problem solving | 2–4 months | Write programs from a blank file |
| Data, SQL and statistics | 1–2 months | Clean and analyse a real dataset |
| Machine learning | 3–4 months | Train, evaluate and deploy classical models |
| Deep learning | 2–3 months | Build CNNs and sequence models in PyTorch |
| Transformers and LLMs | 2–4 months | Fine-tune a model; build RAG and agents |
| Deployment and production | 1–2 months | Ship and monitor an AI app |
*Assumes about 8–12 hours a week. More time per week shortens it; gaps lengthen it.
What speeds you up?
- A fixed time every weekday, so practice becomes a habit.
- A syllabus in the right order, so you never learn a topic before its prerequisites.
- Building something every week, not just watching.
- Getting unstuck quickly: a mentor or doubt session saves days.
What slows you down?
- Jumping between courses and tutorials without finishing any.
- Skipping programming or maths, then struggling with every later topic.
- Long breaks: momentum is hard to rebuild.
Why Program Zero is 18 months
Program Zero takes 18 months (78 weeks) because it covers all of these stages properly, plus full-stack development and security, at a pace that fits around work or college: 90 minutes live every weeknight. You can see every week's topics on the curriculum page. For a deeper comparison of formats, read live training vs recorded courses; if you work full time, see learning AI while working.