You Don't Really Need Another MOOC

Eugene Yan's argument is simple and a little uncomfortable: most of us don't need another course. We just think we do. MOOCs are great for the first pass at Python, Spark, or machine learning basics — structured lessons, clean exercises, satisfying completion certificates. But once you've got that baseline, a third or fourth specialization on the same topic barely moves the needle. Yan compares it to a geometric series that converges near 2, meaning the returns after the first course or two shrink toward nothing.

The deeper problem, he says, is that watching lecture videos feels like learning but isn't. It's passive. Kids don't learn to ride bikes from tutorials, and engineers don't learn to code from watching someone else type. Worse, MOOCs live entirely on what Yan calls the happy path — code that copy-pastes and runs, data that's already clean, no CUDA errors, no 2 a.m. stack-overflow spirals. That smoothness is exactly why completion rates stay high and exactly why the skills don't transfer. Real data is dirty. Real debugging is required. Real experimentation means failing repeatedly before something works.

His fix is what he calls just-in-time learning, borrowed loosely from manufacturing logic: you learn only what you need, right when a project demands it. You don't need to master Python generators before writing your first script, and you don't need three Spark certifications before touching two million rows of data. Start the project, hit a wall, google the error, learn that one thing, move forward. Yan cites Andrej Karpathy's advice to attack skills depth-first through concrete projects rather than breadth-first through theory, and to compare yourself only to your past self, not to other people's résumés.

He's also blunt about the excuses people give him — not enough time, too many errors, not job-ready yet. His answer to all of them is roughly the same: the friction you're avoiding is the learning. Companies increasingly care less about certificates and more about whether you can actually build something, and interviews increasingly test judgment on messy, open-ended problems rather than trivia a MOOC can drill into you. A weekend project, six to eight weeks long, on a topic you're genuinely curious about, will teach you more than another polished playlist ever will.