Anthropic Oceanus leaks 🤖, ChatGPT Dreaming 💭, recursive self improvement 🚀

# What You Need to Know This Week Anthropic accidentally revealed details about an unreleased AI model called Oceanus that appears to be more powerful than their current offering. Meanwhile, OpenAI is testing a new "dreaming" feature in ChatGPT that lets the AI think through problems privately before responding, similar to how humans might sketch out ideas before presenting them. Both developments suggest AI companies are racing to build smarter, more capable systems that can reason through complex problems more effectively.