My Blogs
September 26, 2026
Do AI agents investigate code history better with a few good tools, or full access?
A pre-registered study testing whether curated tools or an open code-execution sandbox produce better answers from an LLM agent investigating a software repository's history.
August 26, 2026
EARS: An Aspect- and Emotion-Aware Recommender System
How EARS combines emotion classification and aspect-level sentiment extraction to build product recommendations that understand more than star ratings, and what eight failed designs taught me along the way.