🔍 Read the full analysis: The Uneven Edge Of AI-Powered Research on ThorstenMeyerAI.com
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
TL;DR
OpenAI’s latest internal report says coding-agent runtime in its research organization increased from 0.5 to 5.7 eight-hour workdays per researcher per calendar day since its September report. The company describes the figures as partial indicators of AI use, not proof that research has become faster or that OpenAI plans full recursive self-improvement.
OpenAI says coding-agent use in its research organization has risen sharply since the company’s September report, with average agent runtime reaching 5.7 eight-hour workdays per researcher per calendar day, up from 0.5. The internal measurements describe how researchers use AI tools to develop AI systems, but OpenAI says they are a partial account and do not show how much the tools have improved research output, as discussed in the original analysis of research acceleration.
The report measures agent use in several ways. At API list prices, the median researcher used agentic AI tools worth about $2,400 per day, compared with $1,500 in September. At the 90th percentile, estimated daily use was about $23,000, up from $11,000. OpenAI describes those as increases of 62% and 105%, respectively. The spending estimates combine changes in activity, model selection, model pricing and agent software; they are not a direct measure of work completed.
OpenAI defines an “agent workday” as eight hours of agent runtime. On pretraining and posttraining teams, the average was about eight agent workdays per researcher per day; on scaling and infrastructure teams, it was about four. The company says the overall runtime measure generally captures interactive use and may miss automated activity. By mid-October, more than two-thirds of the researchers in the analysis had a daily peak of at least four agents running concurrently, and more than 10% averaged a daily peak of at least 16.
The report also compares estimated agent use with researcher compensation. As of September 2026, OpenAI put the API-equivalent value of agent use at about twice non-equity compensation. That comparison excludes equity, which the company says generally accounts for 70% to 80% of researchers’ target compensation. OpenAI also says internal AI deployment represented 2% of total research compute. These figures describe different kinds of inputs and do not provide a full accounting of research costs.
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.
Agent use is rising
Eight hours of agent use counts as one agent workday.
- 0.5 average agent workdays per researcher per day
- $1,500 median daily agent use
- $11,000 daily use at the 90th percentile
- 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.
More agents work concurrently
By mid-October, many researchers had several agents active at their daily peak.
More than two-thirds
Had a daily peak of at least four agents working at once.
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.
Runtime and spending do not show whether agents produced useful work or improved research pace, quality, or success.
API list prices reflect changes in usage, model prices, and agent behavior; the compensation comparison excludes equity.
The workday measure may miss automated activity, and the report describes its account as partial.
What Higher Agent Use Can Show
The measurements offer a view into how a leading AI company is allocating tools and compute as it develops future systems. Longer agent runtime and more concurrent agents could change how researchers write code, test ideas and manage technical work. That makes the figures relevant to discussions about whether AI is beginning to contribute to the development of later AI systems.
But usage is not the same as impact. The reported spending and runtime do not establish whether agents produced useful results, replaced researcher effort, improved the quality of experiments or sped up progress toward more capable models. OpenAI’s figures are indicators of activity and resource use, not evidence of a measured increase in research productivity.
The company presents measurement as part of a wider conversation about recursive self-improvement—the possibility that AI systems could help accelerate the creation of subsequent systems. OpenAI says tracking relevant indicators could inform discussion of capabilities, risks and safeguards. It also says this report should not be read as an intention to pursue full recursive self-improvement.
Products Worth Considering

Sony WH-1000XM4 Wireless Noise Cancelling Headphones (Silver)

Samsung Odyssey OLED G8 34-Inch Curved Ultra WQHD Gaming Monitor

Logitech MX Mechanical Mini for Mac Wireless Keyboard, Space Gray
As an affiliate, we earn on qualifying purchases.
From September Baseline to October
The update follows an OpenAI report published in September on research acceleration. The company says the new 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 through which inputs are turned into model capabilities, and changes in those capabilities.
The measurements presented here focus mainly on coding agents as one research input. OpenAI says research also depends on data, human labor and compute, none of which is fully captured by the agent-use figures. Its spending comparison uses monthly API-equivalent agent use and monthly non-equity compensation; the labor-cost measure includes salary, overtime, bonuses, benefits and employer taxes, but not equity.
OpenAI frames the report as a contribution to shared practices for measuring AI research and development. It says other organizations have also contributed to this work, with the stated aim of helping the public assess how oversight and safeguards may need to respond as AI capabilities grow.
What the Usage Figures Cannot Resolve
The report does not establish how much agent use has affected the pace, quality or success rate of OpenAI research. It does not show whether longer runtimes led to useful work, freed researchers for other tasks, or changed model capabilities. The source material mentions planned analysis of researcher contributions and time allocation but does not provide those results.
Several measurement limits also remain. The runtime figure generally covers interactive use and may omit automation. API list-price estimates reflect activity as well as changes in models, prices and agent behavior. The compensation comparison excludes equity, and the supplied material does not give enough detail to evaluate how representative the researchers were or how consistently measurements were calculated across teams. OpenAI describes the overall picture as partial.
OpenAI says it intends the report to contribute to shared measurement practices for AI research and development. The next useful evidence would connect tool use with outcomes: how agent contributions are assessed, how researcher time changes, and whether those changes correspond to faster or better research. The supplied material does not give a date for that analysis.
For now, the figures establish a marked rise in reported coding-agent runtime and estimated spending inside the organization, alongside clear limits on what those measures prove. Further reporting and comparable data from other organizations could help show whether the pattern is specific to OpenAI or part of a broader change in AI research.
Key Questions
What did OpenAI report?
OpenAI reported that average coding-agent runtime in its research organization rose from 0.5 to 5.7 eight-hour agent workdays per researcher per calendar day since its September report.
Does the increase show that OpenAI research is faster?
No. The figures measure agent runtime and estimated spending. OpenAI says they do not establish changes in research speed, quality, success rates or model capabilities.
What does an agent workday mean?
OpenAI counts eight hours of agent use as one agent workday. The company says its measure generally covers interactive use and may miss some automated activity.
How much did researchers spend on agent use?
At API list prices, estimated daily use was about $2,400 for the median researcher and about $23,000 at the 90th percentile. These estimates are not a direct measure of research output or total costs.
Does the report mean OpenAI plans full recursive self-improvement?
No. OpenAI says the report’s measurements are intended to inform discussion of AI research and safeguards, and that the update is not an indication that it intends to pursue full recursive self-improvement.
Primary source: OpenAI · via ThorstenMeyerAI.com
