“Karin Vedia observes that Anthropic's models are better at agentic polling tasks, where they will wait and continuously poll for a result, whereas GPT models tend to stop and wait for user input in similar situations.”