For medium tasks, the first answer is often wrong, so thinking further helps it get to the correct answer. With hard tasks (or puzzles in the case of this research), it never manages to figure it out regardless.

→ View original post on X — @thatroblennon