Early science acceleration experiments with GPT-5

An empirical examination of GPT-5's contributions to multidisciplinary research workflows

Estimating AI productivity gains from Claude conversations

Evidence from large-scale interaction data

Adversarial poetry as a universal single-turn jailbreak mechanism in LLMs

Assessing the effectiveness of poetic reformulations in circumventing safety constraints

What does it take to be a good AI research agent? Studying the role of ideation diversity

The role of early-stage ideation diversity in autonomous research performance

ToolOrchestra

Elevating intelligence via efficient model and tool orchestration