A personal, Claude-powered applied math & statistics school. Give it an academic paper (or just name a method/concept), and Claude designs a complete course that teaches you the fundamentals first, then works up to the real thing — with toy numerical examples, plain-English notation decoding, symbol pronunciation tables, interactive Quarto R notebooks, practice problems with hidden answers, and verified links to external videos and posts.

▶️ Browse the live site — every example course, rendered with runnable in-browser R (WebR) cells. No install needed.

This repo ships tailored to one example learner — a strong R programmer whose math stops at rusty AP-Calculus-level algebra and calculus — described in .claude/course-authoring/learner-profile.md. Every course is written for that profile, so before you use it for yourself or someone else:

1. Edit .claude/course-authoring/learner-profile.md to describe the real learner: their programming background, math level, what they already know, what trips them up, and how they like to learn. This one file retargets the whole system.
2. (Optional) tune notation-style.md (bridge every symbol to their fluent language) and resource-curation.md (their preferred videos and sites).

Everything else — the TEACHING.md contract, the skills, the templates, and the example courses — is learner-agnostic and works as-is. Click Use this template (or fork) to start your own. To keep a course you build private, git-ignore its courses/<slug>/ folder and its rows in _quarto.yml.

The three courses under courses/ double as worked examples of the three ways to seed a build: a concept from scratch (simple-linear-regression), a ported library (gradient-descent), and a written request (logistic-regression, from source-materials/logistic-regression/COURSE-REQUEST.md).

1. (Optional) Drop a paper into source-papers/ (see its README for naming).

2. Ask Claude, in this repo:

   Create a course for the methods section of source-papers/2024-smith-mixedmodels.pdf

   or

   Create a course that teaches me what a Kalman filter is

3. Claude runs the create-course skill: it maps out every prerequisite you'll need (down to your actual math baseline), reuses building-block modules from foundations/ where they already exist, writes the new ones, and builds the course under courses/<slug>/.

• Interactively (recommended): open the repo root in RStudio/Positron (not a single course folder) — that's what activates renv, so the Preview button and chunk-running find knitr/rmarkdown. Then open a lesson.qmd and run it chunk-by-chunk, in the order given by syllabus.md. Tweak numbers, rerun chunks, break things — that's the point.

• As a rendered site: the courses are published at https://radlinsky.github.io/professor-claude/ (auto-deployed from main). To build locally instead — the whole repo is one Quarto project, so render from the root:

  quarto render        # builds everything into _site/ — open _site/index.html

• Start every course from its syllabus.md and 00-roadmap.qmd — they tell you which foundations/ modules to take (or skim, if you've done them before) first.

• R 4.5.2 (the version pinned in renv.lock and CI; nearby versions usually work) and Quarto ≥ 1.9 (CI uses 1.9.38).

• One-time setup (renv): dependencies are managed in a project-local library so every course renders reproducibly. Quarto's R engine needs knitr and rmarkdown to run any .qmd (a "no package called 'rmarkdown'" error means setup was skipped). The committed renv.lock records knitr, rmarkdown, and everything else, so restore it — don't re-init. From the repo root, run once:

  Rscript -e 'install.packages("renv", repos="https://cloud.r-project.org")'   # if you don't have renv
  Rscript -e 'renv::restore()'

  After this, opening the repo root in Positron/RStudio auto-activates the renv library via .Rprofile, so the preview button just works.

• Individual lessons auto-install any other R packages they need in their first chunk. With renv active those installs land in the project library; run renv::snapshot() to record them in renv.lock if you want them pinned.

MIT. Fork it, adapt it, teach with it — attribution appreciated.