Polish programmer Przemysław Dębiak narrowly beat OpenAI's AI model in a 10-hour coding contest, proving humans can still outthink machines for now.

AI isn’t ready to replace human coders for debugging, researchers say Even when given access to tools, AI agents can't reliably debug software.