Skip to content

AI-Powered Pipeline to Stress-Test Research Ideas Before PhD Students Spend a Year on Them

Key Ideas

  • Lopez-Lira built an open-source AI pipeline specifically for evaluating, discarding, and improving PhD student research ideas before they invest significant time.
  • The pipeline is designed for his AI and Asset Pricing PhD course, where students must produce an original paper (not a robustness check or simple extension).
  • The tool works with Claude Code but is also compatible with Cursor, Codex, and Antigravity.
  • The approach treats idea evaluation as a structured, automatable process with explicit criteria for rejection and improvement.
  • The open-source repository is publicly available on GitHub.

Summary

Alejandro Lopez-Lira describes an AI-powered pipeline he built for his PhD course on AI and Asset Pricing. The motivation is practical: PhD students often spend a year on a research idea only to discover it is not viable. The pipeline uses AI to evaluate research ideas early, providing structured feedback on whether an idea is worth pursuing, what its weaknesses are, and how it might be improved. Students are required to produce original research papers, so the pipeline serves as a quality filter and idea refinement tool.

The article is brief, functioning primarily as an announcement of the open-source tool with a link to the GitHub repository. The pipeline is designed to be AI-tool-agnostic, working with Claude Code, Cursor, Codex, or Antigravity, making it accessible to researchers with different tool preferences.

Relevance to Economics Research

This directly addresses a major pain point in PhD advising: the high cost of pursuing weak research ideas. The pipeline represents an emerging use case for AI in economics -- not just writing or coding assistance, but structured evaluation of research ideas before significant human capital is invested. It is especially relevant for finance and asset pricing researchers, and the open-source nature makes it immediately usable by other instructors and advisors.