Perplexity released a new memory system, referred to internally as Brain, that would sit beneath several of its core products rather than living inside any single one. Signals point to Brain powering Perplexity Search, the Perplexity Computer agent, and possibly the Comet browser, positioning memory as a shared connective layer instead of a per-product add-on.

What sets the approach apart, based on what has surfaced so far, is transparency. Where most assistant memory operates as an opaque store, Brain looks set to expose the full knowledge base to the user in three forms:

1. Topics sorted into categories
2. The underlying context held behind each topic
3. A navigable 3D map of the connections between them that a person can hover over and explore

The clustering follows the now-familiar second-brain pattern, closer to an Obsidian-style graph than a flat list, where the model gathers information and groups it by subject. Because context is organized by topic and retrieved only when a task calls for it, the amount pulled into any given request appears lower, which would help both speed and recall quality.

The company has spent the year moving from answering questions toward doing work, with the Computer agent arriving in late February alongside a memory upgrade that sharpened recall accuracy. Brain would extend that arc, turning memory into the substrate that lets search, agents, and browsing draw on the same accumulated context.

Perplexity Brain also lands in a crowded field: always-on agents such as Hermes and OpenClaw lean on their own knowledge bases, and rivals from Claude to Notion are building comparable persistent stores. The feature remains hidden for now, but the recent pace of polishing around it suggests a wider rollout may not be far off.

Official announcement