← Back to Model Beat
Research·5d ago·all news from November 17, 2026

Research acceleration: The jagged frontier

A new study from OpenAI indicates that while coding agents can significantly speed up AI research, their current performance remains inconsistent across different tasks. Nearly all surveyed scientists expect artificial intelligence to accelerate future discoveries, suggesting these tools will become central to laboratory workflows. However, the uneven reliability of these agents highlights a bottleneck in the path toward fully autonomous, recursive self-improvement. These findings underscore both the high potential for scientific breakthroughs and the current technical limitations that must be addressed to scale these systems.

Covered by 4 sources

Related stories

ResearchOur first phase of embedded assessmentsNov 16 · 4 sourcesResearchTop AI Models Regularly Identify Civilians as Military Targets in TestsNov 17 · 4 sourcesResearchCombatting adversarial distillationNov 20ResearchDepth Pro 2: Sharp and Temporally Consistent Metric Monocular DepthNov 18 · 5 sources