# OpenAI Says Its Coding Agents Now Complete 46% of Two-to-Four-Hour Research Tasks, Up From 23% in March

The gains are uneven by task type, the measurement method changed partway through, and OpenAI is calling for labs to be required to report indicators of recursive self-improvement.

- Published: 2026-11-19T06:16:23.269Z
- Canonical: https://polylog.news/ai/2026-11-19/openai-says-its-coding-agents-now-complete-46-of-two-to-four
- Publisher: Polylog (AI desk)
- Section: tech
- Sources: [OpenAI News](https://openai.com/index/research-acceleration-jagged-frontier), [Anthropic Research](https://www.anthropic.com/research/investigating-vault-particles-with-claude), [StartupHub.ai](https://www.startuphub.ai/ai-news/ai-research/2026/openai-agents-do-busywork-not-research), [ZenML LLMOps Database](https://www.zenml.io/llmops-database/scaling-coding-agents-for-frontier-ai-research), [Interesting Engineering](https://interestingengineering.com/ai-robotics/anthropic-first-3d-bacterial-vault-map)

OpenAI published an internal measurement of how coding agents are changing its own AI research. According to secondary coverage of the report, agent success on research tasks that take a human two to four hours rose from 23% in March to 46%…

This story is for subscribers. Read it in full at https://polylog.news/ai/2026-11-19/openai-says-its-coding-agents-now-complete-46-of-two-to-four (subscription information: https://polylog.news/pricing).