About VibeFormer

What this is

A written course in mathematics, statistics, machine learning and modern AI, covering 593 lessons across 26 modules. 81 are written so far; the rest are published as an outline so the whole route and its dependency order are visible from the start.

The syllabus floor is the material taught in Indian Institute of Technology data-science programmes, extended outward into deep learning, language models, retrieval, graph learning and AI safety.

How lessons are written

Every lesson follows the same shape:

• Terms are defined before they are used, from a single glossary of 115 entries, so a definition cannot drift between the place it is taught and the place it is referenced.
• Results are derived rather than asserted, with the algebra shown rather than skipped.
• Every formula is then worked on small numbers you can verify by hand. Each numeric example in this site was computed and checked against exact arithmetic before being written down.
• Pitfalls name the specific mistakes that cost people marks and cause production incidents, with what to do instead.
• Code is written from scratch first, usually in NumPy, so the mechanism is visible before a library hides it.

What it will not do

• No videos. Text is faster to skim, searchable, and checkable.
• No accounts, no paywall on the written curriculum, ever.
• No claims without a derivation or a citation, and no pretending a result is stronger than it is.

How it is paid for

Advertising covers hosting. If graded practice, timed mock papers or spaced-repetition tooling get built later, those may be paid products — the written lessons will not be.

Corrections

Mathematical writing contains errors, and this site will have some. If you find one, please report it — a specific correction is more useful than almost anything else you could send.