Perplexity's AI Agent Now Has a Brain That Learns From Its Own Mistakes
Perplexity launched Brain, a memory system for its Computer agent that stores task-level work history instead of personal profile data. It records what the agent tried, which connectors and sources were used, what users corrected, and what failed, then periodically synthesizes this into a context graph and personal LLM wiki that loads before the next task. Each memory links back to its source for traceability. Perplexity says Brain improves correctness, recall, and cost on repeat tasks, though its figures are internal. The system is aimed at recurring work such as research, monitoring, and reporting, where prior attempts matter. Compared with self-hosted alternatives like OpenClaw and Hermes, Brain is centralized and runs inside Perplexity’s ecosystem, so users get transparency but not full data ownership. It improves performance on familiar tasks, but does not make the underlying model generally smarter or solve cross-domain transfer. Brain is in research preview for Max and Enterprise Max users.
