In a key benchmark, Claude Sonnet 4.5, a next-generation language model, runs autonomously for 30 hours on a single task.
Some call it “vibe-coding” because it encourages an AI coding assistant to do the grunt work as human software developers ...