Today, a new memory system called Brain has been launched for the Computer platform, aimed at enhancing AI capabilities through continuous learning. Brain constructs a context graph from tasks performed by Computer, allowing it to remember past actions and improve its efficiency over time. Early measurements show that this system increases answer correctness by 25% and reduces costs by 13% for tasks needing historical context. This development aligns with broader trends in AI, where the focus is shifting towards creating proactive systems that can identify opportunities and issues as they arise in user workflows, making the agents more effective at delivering results.

Brain: Brain is a self-improving memory system for AI agents that constructs a living context graph from tasks, sessions, corrections, and sources. It reviews accumulated work at intervals to learn what succeeds or fails, enabling agents to perform better over time rather than starting from scratch. This release makes Brain available as a research preview integrated with Computer for Perplexity Max subscribers.
Computer: Computer is Perplexity’s AI agent platform designed for performing tasks with access to connectors, files, and sources. It now incorporates Brain to maintain persistent context across sessions, making each task more stateful and informed by prior projects and decisions. The system uses the resulting memory links for transparency and improved handling of context-dependent work.

`json
{
“AI Agent Memory”: “Developers are enhancing AI memory systems by focusing on agent actions and outcomes, enabling recursive self-improvement in task performance.”,
“Proactive AI Systems”: “Continuous learning is crucial for AI agents to proactively identify opportunities and issues in user workflows and organizational settings.”
}
`