Announced by OpenAI research scientist Alexander Wei on X, the AI scored 35 out of 42 points by solving five of the competition’s six notoriously tough problems.