OpenAI coding Agents now handle 46% of 2-to-4-hour research tasks autonomously
11/18/2026 — 11/22, 03:25·4 sources (all)·4 reports (all)
Key takeaways
- OpenAI reports its coding Agents can independently finish 46% of 2-to-4-hour research tasks, up from 23% in March, while completion for 4-to…
Background & analysis
On 2026-11-18, OpenAI News released a study titled Research acceleration: The jagged frontier, measuring how coding agents speed up AI research. OpenAI said the agents' performance remains uneven across different parts of the research process and discussed what this means for recursive self-improvement. The report was the second in a series on AI accelerating OpenAI's own research, following a first report published in September.
On 2026-11-19, OpenAI shared more specific figures. It said its agents now complete 46% of the tasks that take its researchers two to four hours entirely on their own, up from 23% in March. The same day, The Decoder reported that OpenAI had published its second internal report on research acceleration, noting a sharp rise in usage of internal coding agents such as Codex and Dots. Measured at API list prices, the median researcher consumed tokens worth $2,400 per day in October, a 62% month-over-month increase, while the top 10% of users reached $23,000 per day. The report said these agents effectively deliver 5.7 workdays of output per person per calendar day, while the models still produce large amounts of low-quality content beyond routine tasks.
On 2026-11-22, The Next Web added figures from the same report: with no human help, the agents' completion rate for tasks taking four to eight hours rose from 13% to 35%. The report also recommended that labs publicly track signs of AI accelerating their own research and disclose that researchers' average daily agent costs had risen noticeably compared with September. Public reporting has so far stopped at the release of the second report, with its completion-rate and cost data as the latest available information.
AI-generated from 4 reports · updated 7 hours ago
Latest turnOpenAI says its coding agents now complete 46% of research tasks that would take a human two to four hours without help, up from 23% in March, while success on four-to-eight-hour tasks rose from 13% to 35%, The Next Web reports. The Wednesday report also calls for labs to publicly track signs of AI speeding up AI research itself, and notes researcher agent spend rose sharply from September.
- Heat index
- 78
- All sources
- 4
- All reports
- 4
- First seen
- 4 days ago
Reports on this story headlines open the original
OpenAI says its coding agents now complete 46% of research tasks that would take a human two to four hours without help, up from 23% in March, while success on four-to-eight-hour tasks rose from 13% to 35%, The Next Web reports. The Wednesday report also calls for labs to publicly track signs of AI speeding up AI research itself, and notes researcher agent spend rose sharply from September.
The Next WebAI score 64
OpenAI published a second report on internal research acceleration, saying usage of coding agents like Codex and Dots has surged: at API list prices the median researcher consumed $2,400 of tokens per day in October, up 62% month over month, with the top 10% reaching $23,000 a day. The report says agents run the equivalent of 5.7 workdays per researcher per calendar day, while models still produce plenty of low-quality output beyond routine tasks.
The DecoderAI score 65OpenAI says its agents now complete 46% of its researchers' 2-to-4-hour tasks on their own, up from 23% in March. The median researcher's agent use has reached $2,400 per day, according to the company.
TechmemeAI score 83
OpenAI published research measuring how much coding agents accelerate AI research, finding that their performance remains uneven across tasks. The work also discusses what that unevenness signals about recursive self-improvement.
OpenAI NewsFirst-partyAI score 53
Other stories people are talking about
- Heat index 47Guide: spotting when AI coding agents weaken or delete your tests1 sources
- Heat index 1143RisingAnthropic Reaches $1.6 Trillion Market Value in First Trading Week After IPO21 sources
- Heat index 1011Anthropic Mathematician Says Claude Formalized the Generalized Riemann Hypothesis21 sources
- Heat index 622Anthropic IPO total raise reaches $93.5 billion after greenshoe exercise20 sources
- Heat index 596RisingThinking Machines Lab Open-Sources Inkling 2 Mini with 32GB Model Card Recommendation10 sources
How is heat calculated?About the methodHide
Heat counts how many independent sources covered a story in the last 48 hours: one source counts once no matter how many posts it published, decaying with a 24-hour half-life. What ranks first is what many people are talking about.
This page aggregates public feeds. Headlines and summaries are machine-organized and remain the property of the original authors; verify important facts at the source. If you believe a headline or summary infringes your rights, email the contact address in the footer with the page URL and basis of your claim, and we will remove or replace it after review.
- Surge
- Discussion rising fast
- New
- First report within 6 hours
- Rising
- Still gathering discussion

