Inside OpenAI's agent-powered research boom

OpenAI has released internal data demonstrating the scale of its coding agent adoption across the research organization. Agents are logging 3.1 workdays of work for every human workday invested, with researchers now spending an average of $600+ per day on agent tokens, and the 90th percentile exceeding $7,000 daily. Token output has increased 124x since December, and approximately 80% of researchers now use four or more agents simultaneously. The company has achieved Sam Altman's September timeline goal of creating an "automated research intern," with plans to reach full autonomous researcher capabilities by March 2028. Experiment counts per researcher have reached all-time highs, and agents are successfully handling increasingly complex and longer-duration tasks, with particular success in areas like technical troubleshooting.

In medical breakthroughs, Insilico Medicine published trial data for rentosertib, an AI-designed drug targeting idiopathic pulmonary fibrosis. The drug demonstrated biological age reduction across six independent aging clocks in a 42-patient trial, with one measurement showing patients registering 2.7-3.5 years younger. The optimal dosing for the aging effect differed from the dosing that maximized lung-function improvements, suggesting effects beyond disease treatment alone.

Public sentiment toward AI remains predominantly negative despite rising usage. An NBC News poll of 7,105 adults found 70% feel more worried than excited about AI, with concern spanning both major political parties. While 52% of Americans now use AI very often or sometimes (up 6 points from June 2025), only 18% trust AI-generated information most or almost all of the time. Data centers have emerged as a flashpoint issue, with 69% opposed to local data center construction. Voters express skepticism across multiple domains: 70% believe AI is costing jobs, majorities expect harm to schools and elections, and 81% believe current Washington AI regulations are insufficient. When asked which political party to trust on AI policy, 44% said neither, with low confidence in both Democrats (20%) and Republicans (16%).