Blog
Perspectives, comparisons and case studies from data science and analytics.
Why Your LLM Can't Count the R's in 'Strawberry'
Ask ChatGPT how many R's are in strawberry and it often gets it wrong. The reason isn't a bug — it's tokenization. Here's what's actually happening, with an interactive demo.
Read →An embedding is just a vector
"It captures the meaning of a word" tells you nothing you can act on. The mechanical version: a row of a matrix, looked up by an integer.
Read →What PCA is actually doing to your data
Most explanations stop at "it reduces dimensions." Here is the mechanism, the feature-selection confusion, and the cost nobody mentions.
Read →Your Correlation Matrix Is Not Feature Selection
Dropping one of every pair above 0.8 is the most common ritual in applied regression — and it misses the multicollinearity that actually breaks your model.
Read →Looking for the lessons themselves? Browse the Open Library — free, bite-sized Data Science lessons.