2026
An executable CAD benchmark for language-model CAD agents, with 17 Build123D tasks, difficulty-weighted scoring, model harnesses, and a public results leaderboard.
2026
An execution-based EDA benchmark for agents that reconstruct KiCad PCB projects, combining frozen task packs with ngspice-driven I/O simulation graders.
2026
A research project on residual controllers for reasoning switches, exploring how controllers can steer when reasoning systems change strategies.
Dec 2025
An infinite canvas web browser built with Electron and Typescript.
June 2025
A tool to automate context gathering and applying changes when pair programming with online LLMs.
May 2025
Sign in with your Google account and chat with RAG over your docs.
June 2023
A model built off of Falcon 40B that generates music in ABC notation.
March 2021
$55 for a Linux desktop on your glasses.