Member-only story
Everyone Told Me to Learn Julia. Here’s What I Found After Looking at the Numbers.
Julia is faster than Python and almost nobody uses it for AI. I wanted to understand why, and whether you should learn it anyway.
I bookmarked “learn Julia” for the third year running last month. Same tab, same intention, same quiet guilt. Then a comment on one of my posts finally got to me: “You’re an ML engineer and you still haven’t picked up Julia? It’s 10x faster than Python.” So I stopped bookmarking and started reading the actual adoption numbers. What I found talked me out of the exact thing I’d been feeling guilty about.
Here’s why this keeps coming up now. AI workloads have shifted. We moved from pure cloud-scale training toward small models, local inference, agent loops, and a louder scientific-ML crowd that cares about speed in a way the LLM crowd mostly doesn’t. Every few weeks someone points out that Julia runs circles around Python on numerical code, and they’re not wrong. So I wanted the honest version: what is Julia actually solving, how do you tell if it solves a problem you have, and should you spend real hours learning it.
