Claude Code produces a volatility research paper

Mingze Gao demonstrates an empirical project about the idiosyncratic-volatility puzzle in US stocks from 1990 through 2024. He supplies the question, four standard WRDS extracts and writing rules, including a prohibition on causal language for a descriptive paper.

Claude Code's skills and specialist agents work through literature, strategy, analysis and writing. The reported outputs include an annotated bibliography, three prespecified hypotheses, nine robustness checks, an R script, six tables, two figures and a compiled thirty-five-page paper. Data preparation and manuscript values remain tied to the generated analysis rather than being detached prose estimates.

Paired reviewers catch specific problems: a fabricated citation, a causal verb in the descriptive discussion and a body-text statistic that disagrees with its table after rounding. The public talk links the compiled paper, but these reported checks do not constitute independent peer review or proof of every result. The fifteen-minute title describes the talk, not a measured total project duration. The useful workflow gives specialists written contracts, gates stage progression through review and retains human ownership of the question, defensible methods and final claim. This entry records the research process rather than treating its findings as investment advice.

Prompt
Suggested prompt (written by Agents In Use, not the author's)

Source: Blog · Mingze Gao