Laude Institute

Suite of real-world command-line tasks executed inside sandboxed Docker containers to measure agentic shell competence.

Hard subset of Terminal-Bench - long-horizon shell + filesystem tasks.

70+ competitive text games (negotiation, social deduction, language puzzles) where models play head-to-head and a TrueSkill rating is fit.