OpenAI says it has reached its "automated research intern" milestone, with AI agents performing research work at a 3.1-to-1 ratio compared with human effort inside the organization. The company said its most prolific researchers consume more than $7,000 worth of tokens daily through AI coding agents, based on API pricing rather than OpenAI's actual spending. Median daily token usage in the research organization exceeded $600 by mid-August, up from $162 in July, while users at the 90th percentile exceeded $7,000. OpenAI said agents help researchers ship code faster, run more experiments, delegate increasingly senior-level work and troubleshoot technical issues. The company is targeting a fully automated AI researcher by March 2028. The figures come as companies question whether heavy AI usage delivers sufficient productivity gains. Anthropic holds an 83.5% YES probability in a prediction market for leading the best AI model at the end of September 2026, though that market does not establish which model is objectively best.