Is Claude on the cusp of taking over its own development? What experts saw inside Anthropic

Meter's exclusive analysis of Anthropic reveals that the company has reached a recursive self-improvement threshold where internal AI systems drove an approximate 1.97x acceleration in model development by October 2025. This shift has prompted leading AI safety experts to revise superintelligence timelines to within one to two years, while simultaneously sparking bipartisan legislative calls to ban fully automated research loops. Despite Anthropic's proposed safety monitoring and acknowledgment of crossing its internal speedup line, critics warn that voluntary audits lack the mandatory independent oversight needed to manage the risks of exponentially compounding progress.