OpenAI’s GPT-5.5 has posted an 82.7% score on Terminal-Bench 2.0, a benchmark that throws AI agents into difficult, ...