Julius AI

It does real computation — uploads a file, writes and executes Python, shows the code — rather than hallucinating numbers, which separates it from a general chatbot for statistics. Students who cannot yet write pandas can still reach a regression and spend class time on interpretation. The visible code makes it teachable: you can point at the line that dropped the nulls.

In intro statistics, having students produce a regression and grading only the written interpretation of residual plots and p-values. Giving the whole class the same messy dataset and comparing how different sessions cleaned it, teaching that analytic choices are choices. Having students audit the generated Python line by line and annotate what each step assumes.

No education pricing and a $20/month entry point make classwide adoption an equity problem; free credits will not carry a semester. It makes silent analytic decisions — dropping rows, choosing a default test, imputing — and reports results with full confidence, so it teaches statistical practice badly unless you build the audit step in explicitly. It is not a symbolic math tool; do not expect it for calculus or proof.