I asked Claude, ChatGPT, and Gemini to debug a Python error, and the difference was too noticeable to ignore.
I ditched my terminal for Claude's built-in code executor, and I'm not going back.
Researchers who found the bug warn that its Moderate rating understates a threat reaching across LLM gateways, MCP servers ...