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

🔍 Read the full analysis: Where AI Speeds Up Research—and Where The Frontier Gets Jagged 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’s latest internal report says coding-agent use by its researchers has risen sharply since a September report. Average agent runtime reached 5.7 eight-hour workdays per researcher per calendar day, but the company says the figures measure activity, not research output or the pace of AI progress.

OpenAI says its researchers are using coding agents substantially more than they did in the company’s September analysis, with average runtime rising to 5.7 eight-hour workdays per researcher per calendar day, from 0.5. The internal measurements offer a partial view of AI’s role in developing AI systems, but do not establish that research is progressing faster or that agents are improving results.

The report uses “agent workdays” to express how long coding agents run: eight hours of agent use counts as one workday. OpenAI says the overall average was 5.7 workdays per researcher per calendar day. Pretraining and posttraining teams averaged about eight workdays, while scaling and infrastructure teams averaged about four. The measure generally captures interactive use and may omit some automated activity.

OpenAI also estimates agent use at API list prices. The median researcher’s daily use was valued at about $2,400, compared with $1,500 in September; at the 90th percentile, the estimate was about $23,000, compared with $11,000. The company describes those differences as increases of 62% and 105%, respectively. It cautions that the estimates reflect several factors, including usage levels, model choices, and changes to models and agent software—not a direct tally of productive work.

By mid-October, more than two-thirds of researchers in the analysis had a daily peak of at least four agents running concurrently, OpenAI says. More than 10% averaged at least 16 agents at their daily peak. The company also says internal AI deployment accounts for 2% of its total research compute. These figures describe use and resource allocation; the report does not show how much agent activity contributed to new research results.

At a glance
reportWhen: Latest report; compares measurements wi…
The developmentOpenAI has released new internal measurements showing increased coding-agent use across its research organization, including higher runtimes and API-list-price spending than in its September analysis.
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 measurements matter because AI systems are increasingly being used in work that may contribute to building later AI systems. If agents help researchers write code, test ideas, or manage technical tasks, their use could affect how research teams work. OpenAI’s data show that agent use within its own research organization has expanded, but runtime and spending are inputs, not evidence of faster progress.

OpenAI frames the report as part of a wider effort to measure AI’s contribution to research and development. The company argues that shared indicators could help inform public discussion about capabilities, risks, and safeguards. Such reporting may be useful to researchers and policymakers, but the figures are specific to OpenAI’s organization and methods; they do not, on their own, establish a trend across the AI industry.

The distinction is particularly relevant to discussion of recursive self-improvement, in which AI could help accelerate the development of later AI systems. OpenAI says the report should not be read as an indication that it intends to pursue full recursive self-improvement. The reported rise in agent activity describes current tool use, not a plan or proof of an autonomous cycle of AI development.

Products Worth Considering

LG 27GS60QC-B Ultragear 27-inch Curved Gaming Monitor (QHD, 180Hz, 1ms, 1000R)

LG 27GS60QC-B Ultragear 27-inch Curved Gaming Monitor (QHD, 180Hz, 1ms, 1000R)

DJI Mic 3 Wireless Microphone (2 Transmitters, 1 Receiver, Charging Case)

DJI Mic 3 Wireless Microphone (2 Transmitters, 1 Receiver, Charging Case)

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.

How the September Baseline Fits

The new report follows an OpenAI analysis published in September on research acceleration. The company says this update adds quantitative and qualitative measures of AI’s contribution across research areas and of continued human involvement. It organizes the topic around research inputs, the process of turning those inputs into model capabilities, and changes in those capabilities. The agent figures presented here focus mainly on coding agents as one research input.

OpenAI says research also depends on data, human labor, and compute. Its agent-use statistics do not measure all of those inputs or how they interact. The company compares monthly API-equivalent agent use with monthly non-equity researcher compensation. That compensation measure includes salary, overtime, bonuses, benefits, and employer taxes, but excludes equity.

The compensation comparison needs that caveat: OpenAI says equity generally makes up 70% to 80% of researchers’ target compensation. The company estimated that, as of September 2026, the API-equivalent value of agent use was about twice non-equity compensation. That ratio is not a full comparison of agent costs with total researcher compensation, and the report also says internal AI deployment represents 2% of total research compute.

What the Activity Figures Leave Open

The report does not establish whether greater agent use has changed the speed, quality, or success rate of research. It does not show whether agents produced useful work, replaced tasks previously done by researchers, or contributed to improved model capabilities. Spending at API list prices is also affected by usage, model prices, and software changes, so it is not a direct measure of work completed.

The scope and methods leave further questions. OpenAI says the runtime figure generally covers interactive use and may miss some automated activity. The supplied material does not give enough detail to judge how representative the researchers included are or how the measures were calculated across teams. It refers to planned analysis of researcher contributions and time allocation, but does not provide those results here.

The compensation comparison excludes equity, and the company describes its account as partial. The report’s figures therefore do not provide a complete view of research costs, productivity, or AI’s effect on progress. They also do not establish that the increases observed inside OpenAI apply to other organizations.

Further Measurement and Reporting

OpenAI says it wants the report to contribute to shared measurement practices for AI research and development. The company’s stated aim is to make clearer reporting available as AI systems become more capable, so the public can consider how oversight and safeguards may need to change.

The next useful evidence would connect agent activity to research outcomes: what tasks agents perform, how researchers’ time changes, and whether the work affects the pace or quality of model development. OpenAI’s report points to planned analysis of researcher contributions and time allocation, but the supplied material does not give a publication date or results for that work. Until those measures are available, the figures are best understood as indicators of how intensively agents are used—not proof of the effect on research progress.

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 analysis. Its report also gives estimates of agent spending and concurrent use.

Does the report show that AI has made OpenAI’s research faster?

No. The figures describe agent runtime and estimated spending, but do not establish changes in research speed, quality, or success rates.

How does OpenAI calculate an agent workday?

The company counts eight hours of agent use as one workday. It says the measure generally covers interactive use and may miss some automated activity.

Does the report mean OpenAI is pursuing recursive self-improvement?

OpenAI says the update is not an indication that it intends to pursue full recursive self-improvement. The report describes current use of coding agents in its research organization.

What information is still missing?

The report does not show how much agent use changed research outcomes, whether agents displaced or supported particular researcher tasks, or how representative the measured researchers are. OpenAI refers to further analysis of researcher contributions and time allocation, but the supplied material gives no results or release date.

Primary source: OpenAI · via ThorstenMeyerAI.com

You May Also Like

We Asked Grok AI Whether Ethereum Is A Buy Under $2,400

A 24/7 Wall St. headline says it asked Grok whether Ethereum is a buy below $2,400, but the answer and supporting analysis are unavailable.

A Step-by-Step Look At Warp And MjWarp For Robotics Workflows

Hugging Face shows how to prepare an SO-101 robot simulation in MJWarp across up to 2,048 environments, without reporting a speed benchmark.

What Radisson Hotel Group’s ChatGPT Hotel Discovery Means For Travelers

OpenAI’s announcement says Radisson hotel discovery is coming to ChatGPT, but launch timing, features and booking options remain unspecified.

Exploring How ByteDance, Runway, And Luma Are Reshaping AI Video Trends

ByteDance, Runway, and Luma now reportedly dominate the AI video market, leaving OpenAI behind, though specific market data remains undisclosed.