Anthropic has launched Claude Opus 5, a frontier model featuring a 1-million-token context window that can self-verify and recover from errors without human intervention. While the model shows increased accuracy on knowledge tests, its hallucination rate has climbed to 50%, marking a growing tension between autonomous capability and consistent reliability in AI-driven software development.

Opus 5 represents a shift where AI models are increasingly capable of autonomous work but require higher vigilance due to their propensity for confident 'hallucinations.' Builders should anticipate that coding execution is no longer a viable competitive advantage and pivot toward non-technical moats like exclusive data or deep domain branding.