A quarter of the work of making the next mind is reported as led by the present one

Around 17 September Anthropic disclosed that its model Claude now "leads" 26% of the company's own AI research and development work, up from under 1% in February, with about thirty thousand agents working simultaneously on its most-used internal research platform in August and every action passing through a monitor, which by the company's account blocked about one decision in 47,000. Every part of this is the company's own measurement, largely produced by Claude, and the coverage reports no outs ...