Sam Bowman

Petri is an open-source alignment auditing agent that enables researchers to test AI safety hypotheses in minutes rather than weeks by autonomously crafting environments and running multi-turn audits against target models.

“Co-author of Petri and Petri 2.0 research papers on AI safety”

Alignment is not a constraint on capable AI systems but rather a core component of capability itself; Anthropic's approach of embedding alignment researchers in capability work has produced superior models like Claude Opus 4.5.

“Stated that alignment researchers are involved in every part of training at Anthropic without a clear split between alignment research and applied finetuning.”

Google's NotebookLM can efficiently generate high-quality infographics and slide decks from academic papers, making it useful for educational and professional presentations.

“Mentioned in context of NotebookLM team responding to his tweet with a gif about AI content.”