AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: How Researchers Can Navigate AI’s Jagged Frontier on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get the latest gadgets delivered free with Prime

  • Fast, free delivery on millions of items
  • Prime Video, Amazon Music and more included
  • Member-only deals all year
Start your free Prime trial Free trial for eligible customers · Cancel anytime
As an affiliate, we earn on qualifying purchases.

TL;DR

OpenAI reports that average coding-agent runtime in its research organization rose from 0.5 to 5.7 eight-hour workdays per researcher per calendar day. The company calls the figures a partial measure of AI use, not proof that research is faster or a signal that it plans full recursive self-improvement.

OpenAI says coding-agent use in its research organization has risen sharply since its September report, with average runtime reaching 5.7 eight-hour workdays per researcher per calendar day, up from 0.5. The company presents the figures as a partial measure of AI’s role in research, not evidence of productivity gains or an intention to pursue full recursive self-improvement.

OpenAI estimates that the median researcher used agentic AI tools worth about $2,400 per day at API list prices; researchers at the 90th percentile used about $23,000 daily. The company compared those amounts with $1,500 and $11,000, respectively, in September, describing the increases as 62% and 105%. It says the changes reflect several factors, including activity, model choice, and changes to models and agent software, so spending is not a direct measure of research output.

The report defines an agent workday as eight hours of agent use. Average use was about eight workdays per researcher per calendar day on pretraining and posttraining teams, and about four on scaling and infrastructure teams. By mid-October, more than two-thirds of researchers in the analysis had a daily peak of at least four agents working concurrently; more than 10% averaged at least 16 agents at their daily peak, according to OpenAI.

OpenAI says the runtime measure generally covers interactive use and may miss some automated activity. It also estimates that, as of September 2026, the API-equivalent value of agent use was about twice non-equity researcher compensation. That compensation measure excludes equity, which the company says generally accounts for 70% to 80% of researchers’ target compensation. OpenAI says internal AI deployment represents 2% of its total research compute.

At a glance
reportWhen: Update published after OpenAI’s Septemb…
The developmentOpenAI has published updated internal measurements showing increased coding-agent use among its researchers since a September report.
AI Research · Briefing

Research Acceleration The Jagged Frontier

OpenAI reports heavier use of coding agents in its research organization. The figures show activity and resource use, but do not establish faster or better research.

5.7
agent workdays per researcher per day, up from 0.5
$2,400
median daily agent use at API list prices
$23,000
daily agent use at the 90th percentile
2%
of total research compute from internal AI deployment

Agent use is rising

Eight hours of agent use counts as one agent workday.

September
  • 0.5 average agent workdays per researcher per day
  • $1,500 median daily agent use
  • $11,000 daily use at the 90th percentile
Latest report
  • 5.7 average agent workdays per researcher per day
  • $2,400 median daily agent use
  • $23,000 daily use at the 90th percentile

Use differs across teams

Average agent runtime varies by research area.

Pretraining and posttrainingAbout 8
Scaling and infrastructureAbout 4

More agents work concurrently

By mid-October, many researchers had several agents active at their daily peak.

Concurrency

More than two-thirds

Had a daily peak of at least four agents working at once.

High peaks

More than 10%

Averaged at least 16 agents at their daily peak.

What the figures leave open

The measures capture inputs to research, not the effects on its outcomes.

1
Activity is not output

Runtime and spending do not show whether agents produced useful work or improved research pace, quality, or success.

2
Costs are partial

API list prices reflect changes in usage, model prices, and agent behavior; the compensation comparison excludes equity.

3
Coverage is incomplete

The workday measure may miss automated activity, and the report describes its account as partial.

ContextOpenAI says the report is not an indication that it intends to pursue full recursive self-improvement.
The takeaway
Coding agents are becoming a substantial part of researchers’ workflows, but comparable measures of research outcomes are needed to show what that use achieves.

What Agent Use Can—and Cannot—Show

The figures offer a view of how one AI lab is allocating AI tools within its research organization. Longer runtimes and more concurrent agents may change how teams test ideas and write or review code. The report does not establish that those changes have made research faster, improved its quality, or led to more successful results.

The measurements also bear on discussion of recursive self-improvement, in which AI systems help accelerate the development of later systems. OpenAI says tracking research inputs and their effects could inform debate about capabilities, risks and safeguards. But its stated purpose for publishing these figures should not be confused with a commitment to pursue full recursive self-improvement.

For readers comparing claims about AI’s effect on research, the distinction matters: tool use is an input, not an outcome. Spending at list prices and hours of agent runtime can indicate activity, but neither shows how much useful work agents completed or how much human effort they replaced.

Products Worth Considering

Sony WH-1000XM6 Wireless Noise-Cancelling Headphones – Midnight Blue

Sony WH-1000XM6 Wireless Noise-Cancelling Headphones – Midnight Blue

As an affiliate, we earn on qualifying purchases.

The September Baseline and Its Limits

The update follows an OpenAI report published in September on research acceleration. The company says the newer report adds quantitative and qualitative measures of AI’s contribution across research areas and of ongoing human involvement. Its framework considers inputs to research, the process of turning those inputs into model capabilities, and changes in capabilities.

The new figures focus mainly on coding agents, one input among others. OpenAI says research also relies on data, human labor and compute, which the agent-use measures do not capture in full. The spending comparison uses monthly API-equivalent agent costs and monthly non-equity compensation; the latter includes salary, overtime, bonuses, benefits and employer taxes, but excludes equity.

OpenAI frames the report as part of an effort to develop shared measurement practices for AI research and development. The company says other organizations have also contributed to work on measuring research acceleration. The supplied material does not identify those organizations or describe their methods.

Productivity Effects Remain Unknown

The report’s usage figures do not establish how much agents changed research pace, quality or success rates. The supplied material does not show whether agents produced useful work, displaced researcher effort, or contributed to changes in model capabilities. OpenAI describes planned analysis of researchers’ contributions and time allocation, but results from that analysis are not included here.

The measures have further limits. Interactive runtime may omit automated activity, while API list-price spending can change with usage, model prices and agent behavior. The compensation comparison excludes equity, and the company describes its overall account as partial. The supplied material also does not give enough detail to judge how representative the researchers included were or how the figures were calculated across teams.

The timing needs care: the material refers to comparisons with a September report and identifies the compensation estimate as of September 2026, but does not specify the publication date of the update. The figures should not be treated as a complete account of research costs or results.

OpenAI’s Planned Measurement Work

OpenAI says it intends the report to contribute to shared norms for measuring AI’s effect on research and development. Its stated approach is to track not only tools and other inputs, but also the research process and resulting changes in capabilities. The supplied material does not give a date for a further report or specify when the planned analysis of researchers’ contributions and time allocation will be published.

Until more outcome data and methodological detail are available, the new figures are best read as indicators of agent use inside OpenAI’s research organization. Whether that growth translates into faster or more effective research remains an open question.

Key Questions

What did OpenAI report about coding-agent use?

OpenAI says average agent runtime reached 5.7 eight-hour workdays per researcher per calendar day, up from 0.5 in its September report. Its measure generally covers interactive use and may omit some automated activity.

Does the report show that AI made research faster?

No. The report gives runtime and spending indicators, but does not establish how much agents changed research speed, quality, success rates or model capabilities.

How much did researchers’ agent use cost?

At API list prices, OpenAI estimated daily use at about $2,400 for the median researcher and about $23,000 for researchers at the 90th percentile. These estimates reflect multiple factors and are not a measure of output.

Does OpenAI say it plans full recursive self-improvement?

No. The company says the report is not an indication that it intends to pursue full recursive self-improvement. It presents measurement as a way to inform discussion of capabilities, risks and safeguards.

Primary source: OpenAI · via ThorstenMeyerAI.com

You May Also Like

What Claude’s 35,000 Designs Could Mean For Allergy Drug Research

Anthropic says Claude produced 35,000 designs for allergy drug research, but the available account gives no evidence of testing or a drug candidate.

Brain-Computer Interfaces: Sci-Fi Tech or Next Big Thing?

Gazing into the future of brain-computer interfaces reveals a revolutionary technology that could transform how we interact with the world—if it lives up to its promise.

What You Need To Know About Baseten’s Integration With Hugging Face For AI

Baseten is now integrated with Hugging Face, enabling developers to access conversational and text-generation models via Hugging Face’s infrastructure.

Why Future Mobility Tech Is Bigger Than Just EVs

Only by exploring future mobility tech’s broader impact can we understand how it’s transforming urban life beyond electric vehicles.