The developers of Terminal-Bench, a benchmark suite for evaluating the performance of autonomous AI agents on real-world terminal-based tasks, have released version 2.0 alongside Harbor, a new ...