OpenAI's GPT-5.5 Challenges Anthropic's Cyber Hype with Matching Performance

The parity between GPT-5.5 and the highly restricted Mythos Preview suggests that frontier AI's cyber offensive capabilities are rapidly becoming accessible to the public. While these models excel at complex tasks like binary reverse engineering for minimal costs, their inability to compromise physical infrastructure signals a critical boundary in current AI autonomy.