Projects

Agent evaluation on one running base. Swap the agent under test and rerun the same dataset.

A Claude Code-based agent with a web UI, IM channels, and a sandboxed runtime. Core contributor.

A research workbench for people and agents: papers, PDF reading, and a knowledge graph. Core contributor.

Turns static lab manuals into interactive Socratic tutors. Co-developer.

A database course platform with agent-generated SQL practice, ER modeling, and B+ tree visualization.

A skill that builds or repairs a repo-local agent harness from the repository in front of it.

A terminal quiz tool. One prompt turns a subject into tailored exercises.

A skill that pulls papers into Markdown for agent-native reading.

A local-first Obsidian plugin that charts focused time by file and folder.

A VS Code tree for reordering Markdown and Typst headings, with block-level tags.

An iOS app that ties map-based social play to low-carbon habits.