OpenAI and Anthropic appear to be relying less on reasoning as newer models get smarter during pretraining. Read more in today's AI Agenda newsletter: https://t.co/mvE320UzNQ— The Information (@theinformation) April 22, 2026