AI Coding Agents for Software Development
Quantitatively evaluating how development processes, tasks, and context affect the performance of AI coding agents.
I'm a master's student aspiring to be a software engineer, interested in AI coding agents and developer infrastructure.
Graduate School of Information Science and Electrical Engineering, Kyushu University
Department of Information Science and Technology
Faculty of Engineering, Kyushu University
Department of Electrical Engineering and Computer Science
Same choice, different luck: 100 versions of you live one job-hunting year. A browser game.
One Nix configuration for the hooks, skills, and permission rules of Claude Code, Codex, and other coding agents.
A safe JSON CLI for Apple Mail automation on macOS.
Tracks execution state for Pi coding agents, such as changes, checks, and blockers, outside the main LLM.
A Claude Code command that turns notes from technical books into sub-agent definitions.
A LINE service that recommends smartphones through conversation and generates PDF catalogs. Built as a team in Fukuoka Mitou; I handled product data integration, search and recommendation text generation, the runtime environment, and about half of the presentation.
Quantitatively evaluating how development processes, tasks, and context affect the performance of AI coding agents.
Evaluating Coding Agents for Refactoring Under Project-Specific Conventions
IEICE Technical Report, vol. 126, no. 125, SS2026-27, pp. 157–162, 2026
Non-peer-reviewed