Perplexity unveils Brain, a self-improving memory system for its AI Computer platform
The new feature builds a persistent context graph that learns and updates continuously, aiming to eliminate the repetitive prompting problem that plagues most AI assistants.
Perplexity has launched Brain, a new memory system designed to help its Computer agents improve from previous work rather than simply remember user preferences.
The system builds a context graph of the work performed by Computer. At set intervals, such as overnight, Brain reviews that graph and updates the agent’s working context so it can handle future tasks more efficiently.
Perplexity said the goal is to move AI memory beyond personalization. Most memory systems focus on the user, including preferences, contacts, work style and recurring instructions. Brain is focused on what the agent did, what worked, what failed and what corrections were made.
That distinction matters because agent memory is becoming a core part of AI product competition. A system that can learn from prior work can reduce repeated setup, avoid dead sources and start new tasks with a clearer sense of what the user is trying to accomplish.
Brain stores this context through what Perplexity describes as a living context graph. The context layer takes the form of an LLM wiki that is automatically loaded into the agent sandbox, giving Computer a map of the user’s projects, files, sources, people and prior sessions.
The wiki is updated incrementally as Brain reviews sessions, connector results, source document changes and user corrections. Perplexity said this gives Computer stronger signals on where to look, what to avoid and how to produce better outputs over time.
Early measurement results cited by the company show Brain increased correctness by 25% on tasks Computer had seen before. Perplexity also said recall improved by 16%, while the cost of tasks requiring historical context fell by 13%.
The launch is part of a broader race to make AI agents more persistent and proactive. Instead of treating each prompt as an isolated request, Brain is designed to let agents build institutional memory around a user or organization’s work.
Perplexity said the feature is rolling out today to Max and Enterprise Max subscribers in Research Preview. The company said more Brain capabilities will be announced later.