{
  "id": 6105124,
  "title": "OpenAI Says Its Team Relies on AI Coding Agents For More Work Hours Everyday Than Human Researchers",
  "url": "https://urgent.news/2026/09/07/openai-says-its-team-relies-on-ai-coding-agents-for-more-work-hours",
  "topic": "ai",
  "section": "AI",
  "published": "2026-09-07T04:42:25.000Z",
  "source": {
    "name": "Free Press Journal",
    "slug": "free-press-journal",
    "url": "https://www.freepressjournal.in/tech/openai-says-its-team-relies-on-ai-coding-agents-for-more-work-hours-everyday-than-human-researchers"
  },
  "original_language": "en",
  "account": "OpenAI has released an assessment on the impact of AI coding agents on the company's research operations, stating that staff now rely on these agents for more work-hours than they personally contribute. As of mid-August, the research team was utilizing the equivalent of 3.1 agent-workdays for every human researcher's workday. This shift was noted to have begun in June. Initially, at the start of the year, researchers were using coding agents sparingly, but by mid-August, the median researcher was spending over $600 a day on inference at API pricing. The top 10th percentile of users were consuming more than $7,000 worth of tokens daily. Usage has grown faster than any other function at OpenAI, with the typical researcher's output increasing 124-fold since December. A significant number of researchers are now running multiple agents concurrently, with one in four juggling four or more sessions. OpenAI attributes this acceleration to agents' ability to expedite various stages of research, from designing improvements and building evaluation tools to testing at scale and catching errors. The company's research output per active contributor has surged and the number of experiments conducted per active researcher reached a peak in August, largely due to the wider adoption of OpenAI's Codex coding tool. The company cautions that while these metrics are easy to track, they do not fully capture research progress and that compute availability, not just agent performance, may become a limiting factor in experiment speed. Agents are now handling more extensive and complex tasks, including deciding priorities, designing experiments, building code, running training jobs, analyzing results, and communicating findings. Growth is observed across all these areas, with writing research and infrastructure code being the most common uses. The company also observed a decrease in traffic to an internal help channel used for troubleshooting, suggesting that agents are taking on a larger share of this work, with some teams even eliminating human-staffed support sessions.",
  "summary": "OpenAI has published a detailed self-assessment of how AI coding agents are changing the pace and shape of research inside the company, revealing that its research staff now collectively rely on agents for more work-hours than they put in themselves. According to the company, as of mid-August, its research organisation was drawing on the equivalent of 3.1 agent-workdays - measured against a…",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}