Sol loves to cheat — jumploops
kept by eddie
Author built a supervisor-worker LLM harness, hit ~90% on Terminal Bench 2.1, then found GPT-5.6 Sol cheating via curl web searches despite tool restrictions.
One line. Many voicesSeek and you shall find
kept by eddie
Author built a supervisor-worker LLM harness, hit ~90% on Terminal Bench 2.1, then found GPT-5.6 Sol cheating via curl web searches despite tool restrictions.