AI at work · course 24 of 105
Machine Learning in Plain Numbers
Follow a machine-learning report or data interview: the key words, precision and recall, overfitting, train/test splits, and three models worked by hand. No code.
Not on sale yet. When sales open: $4.99 USD for this course, paid once, or every course for $9.99 USD a month. Prices
$4.99 USD, paid once. Or every course for $9.99 USD a month until you cancel; try it free for 7 days first (email only, no card, ends on its own). Tax added at checkout where it applies.
Runs in Claude (claude.ai, the Claude apps or Claude Code) with your own Claude account. A course opens there as soon as your payment goes through; nothing is shipped.
Preview
What it teaches
The machine-learning words people nod along to (features, labels, regression vs classification, overfitting, validation and test sets) and the maths behind a model report: accuracy, precision, recall and F1 from a confusion matrix, why accuracy lies on rare classes, MAE, MSE and RMSE, train/test splits and k-fold cross-validation, then three models by hand: k-nearest neighbours, a classification threshold, a linear prediction, and the baseline that always guesses the most common class. No code. The AI asks 3 questions, places you, shows a short card per module, then gives you one problem at a time and checks your answer against keys computed by script, naming the slip when it’s a common one (“that’s recall; precision divides by everything flagged”).
How long
41 items. About 25-60 rounds plus a 12-item final: 2-3 sessions (estimate, not timed with a real student). The tutor gives a one-line save code at the end of each session.
The modules
4 modules, drilled one item at a time.
- M0
The words
Concept questions
- M1
Scoring a classifier
Maths
- M2
Regression error, data splits and fit
Maths
- M3
Models by hand
Maths
Before it teaches, it asks
- 1What brings you here: a job that uses models, a course, an interview, or curiosity? Have you trained or read up on a model before?AI voice (ElevenLabs)
- 2A spam filter checked 200 emails: 40 spam caught (TP), 10 good emails flagged as spam (FP), 20 spam missed (FN), 130 good emails passed (TN). What’s the precision?
- 34-fold cross-validation on 1,200 training examples. In each round, how many examples train the model and how many validate it?
What’s exact, and what isn’t
Show
Straight from the course’s own notes.
- Computed by script
- Every calculation, with exact fractions (the one square root rounded to 2 places); every classification metric cross-checked by building the actual and predicted label lists and counting; k-NN distances computed and checked for ties; 13 of the 23 listed wrong answers reproduced by applying the slip.
- Sourced
- Definitions from Google’s Machine Learning Crash Course and ML Glossary and the scikit-learn user guide; short quotes matched word for word on 2026-10-10.
- Simplified on purpose
- Small made-up numbers so the maths fits on paper; binary classification only.
- Not in it yet
- ROC curves and AUC, log loss, gradient descent, neural networks, writing code.