Anthropic researchers surveyed 1,334 employees and 10,980 US adults in August regarding AI capabilities. By 2030, 91% of staff expect AI to achieve Nobel-level breakthroughs, compared to 39% of the public. Furthermore, 55% of Anthropic employees anticipate AI performing at least 80% of their tasks by the end of the decade.
The company reported a 45% revenue increase in Q3 compared to Q2, reaching $16.8 billion. Anthropic plans to allocate 5% of its shares to a new public benefit fund to support those displaced by automation. Employees estimate that AI currently performs 51% of work on advanced models, with projections rising to 95% by 2030.
Anthropic's economic analysis for 2030 outlines three growth scenarios for the US GDP, which was $30.8 trillion in 2025. In an extreme scenario, the economy could be one-third larger than projected without AI, though white-collar unemployment could reach 17.9%. Conversely, non-office roles might see unemployment drop to 3.9%.
Technical benchmarks show rapid progress: the Claude Opus 5.5 model requires 2.3 times fewer computational resources for scientific planning than human experts. Meanwhile, the company has launched Claude Duet, an AI agent capable of managing schedules and dashboards, though it still requires human authorization for final actions.
