In the first decision-tree lesson, ten seeds asked a yes-or-no question: will this seed sprout? A useful split made the answers on each side less mixed.
Now imagine that the label is not sprouted or did not sprout, but the height of the plant after four weeks. The boundary still divides one group into two. The tree still repeats the same propose–judge–repeat loop. But “less mixed” needs a new meaning, and every leaf must predict a number.
The new interactive lesson lets you reuse the intuition from classification, replace class impurity with squared error, and watch each leaf settle on the mean of the observations it receives.
That is the useful connection between the two articles: a decision tree is not one finished diagram or one special formula. It is a way to search for questions that make the next prediction simpler.