I know the names of the tests, but I struggle to choose the right method depending on whether the goal is comparison, examining relationships, or prediction, and I mistakenly say that something is a cause when the p-value is small. We first look at the data type, then interpret the test statistic and p-value in the following order: chi-square, t-test, ANOVA, correlation, and regression.

This covers the issue of having sales data in Excel while the UI, API, and database are separate, making it impossible to check everything at a single address. Vite, FastAPI, PostgreSQL, and nginx are organized into separate folders, and the code received through vibe coding is verified by line count and the customer-facing screen.

Through the single problem of predicting movie ratings, this connects PCA, SVD, matrix factorization, and attention from the perspective of inner products. People who have only called libraries will be able to explain the meaning of embedding numbers and how transformers compute things on their own. It is a journey toward understanding, in a single formula, the entire process of filling in the blanks in a recommender system.