BREAKING
Perplexity unveils 'Brain' for agents
A memory that learns from work
How Brain works
1
Record task to graph
↓
2
Synthesize overnight
↓
3
Update LLM wiki
↓
4
Load into next task
0
%
Answer correctness
0
%
Recall
0
%
Cost for history tasks
Promise versus concerns
Praise
●
Competitive for agentic work
●
Improves with repeated use
Concerns
●
Heavy token consumption
●
May resemble existing Spaces
Preview for Max and Enterprise
AI NEWS BLITZ
Perplexity has launched Brain, a self-improving memory system for its AI agents.