How Pew Research Center Is – and Is Not – Using AI in Our Work

How Pew Research Center is – and is not – using AI in our work

In this post, we’ll share our current guidelines for the internal use of artificial intelligence and potential areas of experimentation.

“Stateful systems are incredibly hard to build”: How Perplexity thinks about AI agent sandboxes

Sandboxes for AI agents may feel like a solved problem. After all, projects like Firecracker, the open-source microVM technology AWS built to power its own Lambda service, provide strong isolation off the shelf, and a growing roster of startups sells managed sandboxes for agent workloads. Still, Perplexity decided to go back to the drawing board because the team there was convinced that isolation isn’t actually where the hard problems are. On Ju…

Research: How AI Agents Broaden the Scope of Knowledge Work

AI agents don’t just speed up knowledge work; they change its nature. In a field study of Perplexity’s conversational assistant and autonomous agent, researchers found that not only did the agent perform more autonomous work,  cut task time, lowered cost, and left users more satisfied, but  it also expanded the scope of work people attempted—crossing occupational boundaries, tackling more complex, multi-step projects, and generating tasks that r…