Modules

Six sessions, one arc. Modules 1–5 build — each module page is the module’s full student notes, with its figures, exhibits, lab link, and slide deck. Module 6 discusses. The modules assume you read them in order: every module leans on the habits the previous one installed, and the sequencing is deliberate (the context-window mental model precedes everything; skills are learned in Module 3 so Module 4 can apply them; the WRDS checkpoint in Module 2 exists so Module 4 works).

Foundations — Tools, Context, and the Research Map. The tool ladder, the stateless-agent mental model, goal-first prompting, CLAUDE.md, permissions, and the three mantras the course repeats throughout.
The Empirical Loop on SEC FSDS. One accounting question travels from vague idea to verified, publication-style figure — with named scripts, paper trails, and sanity checks as the deliverable.
EDGAR Text → Structured Data, + Your First Skill. A 10-K text pipeline with caching, quality reports, and measurement-validity discipline; then the SOP becomes a reusable SKILL.md.
WRDS MCP, Scale, and the Stata Bridge. The flagship: plain English to a correctly filtered funda extract on your own credentials, landed in Parquet/DuckDB with a metadata table — and inspected before it’s trusted.
Writing, Safety, and the Verification Capstone. The banal/meaningful delegation line, sandboxes and credential hygiene, verification debt, git as the operating system — and the capstone that hardens one of your pipelines.
Seminar: AI, the Production of Knowledge, and the Academic Research Market. Three questions, each harder to sit with than the last, five short readings, one position memo: “What I will still be paid for in 2036.”