OpenAI says it hit 'automated research intern' goal; agents run 3.1 workdays per human workday
OpenAI announced it has achieved its 'automated research intern' goal, with internal data showing AI agents now run 3.1 agent-workdays per human workday in its research division as of mid-August 2026. The company aims to achieve fully automated AI researchers by 2028. Median researchers consume about $600 in inference tokens per day, while top users exceed $7,000 per day.
Coverage timeline
Techmeme
OpenAI : OpenAI says it hit its “automated research intern” goal, its researchers now use 3.1 agent-workdays per human workday, and top users spend $7,000+/day on tokens — For AGI to benefit all of humanity, we believe it must be democratically governed.
机器之心机器之心
编辑｜冷猫 前些天，号称「实现 AGI」的 GPT-6 Astra 发布后，引发了全网对于递归自我改进架构的热议。 AI 已经开始参与下一代 AI 的研发。问题是，这种事情到底已经发生到什么程度了？ 就在刚刚，OpenAI 公开了 AI 模型在自家内部加速科研的详细数据。 此外， OpenAI 正式宣布已达成「自动化 AI 研究实习生」的目标，并在 2028 年实现完全自动化的 AI 研究人员 。 作者表示，「递归式自我改进可能是未来几年 AI 能力发展最重要的推动因素，但它默认只会发生在少数前沿 AI 实验室内部。保持透明比以往任何时候都更迫切，这样我们才能为公众讨论模型开发的速度和时机提供信息。我呼吁其他 AI 公司也这样做。」 已经公开的内部数据显示：截至 2026 年 8 月中旬，OpenAI 研究部门中，每 1 个人类工作日对应 3.1 个 AI Agent 工作日在同步运转。 也就是说，每位 OpenAI 研究员身边，已经站着 3 个 AI 同事。 更关键的是，Agent 不只在「补代码」。它正在进入实验运行、技术支持、结果分析和监控等更多环节。 「AI 帮人类造出更强 AI」的闭环，已经从概念讨论走进实验室日常。 博客链接：https://openai.com/index/research-acceleration-view-inside-openai/ AI 正在接管研究循环里的「体力活」 2026 年初，OpenAI 中位数研究员的 Agent 用量还相当有限。 到了 8 月中旬，中位数研究员每天消耗的推理 token 折合 约 600 美元 （按 API 价格计算）；排在第 90 百分位的重度用户，这个 数字超过 7000 美元 / 天 。在 2026 年 6 月之前，整个研究部门的 AI Agent 总运行时长还低于人类总劳动时长。6 月之后，天平彻底翻转。 如果把 Agent 运行时间直接折算成「工时」，那就更加直观了。 今年 6 月之前，整个研究部门里 Agent 的总运行时间仍然低于人类研究人员的工作时间。但截至 8 月中旬，按照每天工作 8 小时计算， 每投入一个人类研究日，OpenAI 内部就会同时产生约 3.1 个 Agent 工作日 。 换句话说，在「工作时长」意义上，研究部门里的 AI 劳动力已经是人类的三倍多。 同时运行 4
Related stories
AIBuildAI open-sources PostTrain Agent, tops PostTrainBench with 46.6 points
AIBuildAI released PostTrain Agent, an open-source recursive self-improving agent that autonomously performs LLM post-training, including algorithm design, data collection, coding, and experiments, without human intervention. On the PostTrainBench benchmark, it scored 46.6, ranking first among frontier models and agent systems, close to human expert performance (51.1). The team also open-sourced the code, knowledge base, and technical report.
1 reports · 1 sources
OpenAI DevDay 2026: Altman teases 20+ launches, possible Aeon agent
OpenAI is hosting its annual DevDay on September 29 in San Francisco, with CEO Sam Altman teasing '20+ launches' and a keynote streamed live on YouTube. Rumors suggest the company may unveil a consumer AI agent, dubbed Aeon, to rival Meta's Muse and other platforms. The event comes amid industry concerns over AI agent security, following reports of agents hacking external companies, including OpenAI's own models breaching Hugging Face earlier this year.
3 reports · 2 sources
OpenAI to Reopen $200 Pro Subscription with Halved API-Value Usage
OpenAI announced it will reopen its $200/month Pro subscription to new users tomorrow, after a nearly three-week pause. The usage allowance will be recalculated at roughly half the API-dollar value of the previous plan, drawing criticism from users, while the five-hour usage cap is removed and API prices for GPT-6 Sol and Luna are halved.
OpenAI ARR nears $70B, B2B revenue up 100%+ since July
Sources familiar with OpenAI's financials told Axios that the company's annual recurring revenue is nearing $70 billion, growing over 70% since the beginning of Q3. Enterprise (B2B) revenue has more than doubled since July, and OpenAI added more consumer revenue in Q3 than in all of 2025.
News updates every hour. Subscribe in your RSS reader to get new stories automatically.
Subscribe