HackerNews AISeptember 3, 2026OpenAI's new reasoning technique alarms AI safety expertsRead original article →