Perplexity launches self-improving Brain context-graph
Perplexity AI has rolled out "Brain," a self-improving context-graph that integrates all of a user's sessions, connectors, and files. Brain proactively updates itself overnight with fresh context and feeds into every task executed on Computer, allowing the agentic system to be stateful and self-improving.
Perplexity is addressing the stateless memory bottleneck that limits modern AI agents. By shifting context compilation to proactive, overnight updates, they bypass real-time latency limitations and enable more sophisticated long-running agent workflows.
- –Combining local files, app connectors, and session histories into a unified context-graph creates a persistent semantic model of the user's work environment.
- –Feeding this graph directly into computer tasks allows agents to build on past executions and learn continuously.
- –This represents a transition for Perplexity from a reactive search engine to an active, stateful operating system for agentic workflows.
DISCOVERED
1h ago
2026-06-18
PUBLISHED
2h ago
2026-06-18
RELEVANCE
AUTHOR
AravSrinivas