OpenAI agents log 3.1 workdays per human, hits intern goal

OpenAI's internal report, covered in The Rundown's September 8 newsletter, offers a look at the resources behind its push toward automated research. The figures point to a potential competitive advantage for frontier labs: early access to powerful models gives them time to develop research workflows, while large inference budgets let them put those workflows into practice at scale. The report also shows why the size of that advantage remains difficult to measure.

The report's own numbers carry caveats. The 3.1 agent-workdays figure measures runtime, not productive output, and the researcher population includes support roles. Monitoring overhead adds another cost: OpenAI estimated it at roughly 20% of the inference compute being monitored, with substantial variation. An August 18 disclosure revealed a two-week pause in reinforcement learning training to strengthen safeguards.

The competitive payoff likely hinges on how reliably labs convert all that activity into useful research. Public access to GPT-6 Astra may spread capability without immediately erasing experience accumulated inside a lab — researchers who have already learned to assign tasks, preserve context, and manage interventions may hold a practical head start. Privileged access and deep budgets could reinforce one another, but the report provides no measured lead over rivals or quantified rate of compounding.