OpenAI says it has reached its “automated research intern” goal, according to a Techmeme item summarizing an OpenAI post. The company also says its researchers now use 3.1 agent-workdays per human workday. The same item says OpenAI’s top users spend more than $7,000 per day on tokens. The summary does not specify which users, what workloads those token bills cover, or how OpenAI defines an “agent-workday.” The framing matters because OpenAI is presenting agent use not just as a product feature, but as a measurable input to research work. Still, the available evidence here is limited: the cluster contains one reputable aggregation of OpenAI’s own claims, with no independent confirmation or methodological detail. For now, the clean read is that OpenAI is signaling increased internal reliance on agents and meaningful spend from its heaviest token users. The stronger claims — how much useful research labor the agents are actually performing, and how generalizable that usage is outside OpenAI — are not established by the provided material. Who benefits: OpenAI benefits if the claim strengthens confidence that agentic systems are becoming useful in real research workflows. Heavy users with budgets for large token consumption may also benefit if the agents can absorb meaningful work. Who's exposed: Teams comparing themselves against OpenAI’s reported agent usage are exposed to a measurement problem: the source does not define “agent-workday” in the provided summary. Organizations with lower token budgets may also find the reported top-user spend hard to replicate.