In a field where new models and tools appear weekly, the instinct to "skill up" by taking another MOOC is strong. Eugene Yan's latest essay argues that instinct is often procrastination. He contends that learners should stop accumulating courses and instead start building things, even if the projects are messy and imperfect.
Yan's core advice is to learn just-in-time rather than just-in-case. Instead of pre-studying a curriculum, wait until a real project forces you to need a concept, then learn it deeply enough to apply it. He also encourages going "off the happy path" — abandoning the smooth, well-documented tutorial route to wrestle with broken code, edge cases, and ugly real-world data.
For AI and ML practitioners specifically, the essay is a useful corrective to the "course-completion" mindset. Because the ecosystem shifts quickly, a course finished months later may already be outdated. The essay is an opinion piece, not empirical research, but it aligns with a growing emphasis on portfolio work and applied learning over certificates.