Introducing Brain in Computer. Brain is a continuously learning memory system. Every task on Computer plugs into a context graph built by Brain. It makes Computer more stateful with every run. Available as a research preview for all Perplexity Max subscribers.

With Brain, Computer starts each task with full context of your projects, decisions, and sources instead of from scratch. On tasks that require past context, Brain improves answer correctness by 25%, recall by 16%, and runs 13% cheaper per task.

Every memory links back to the session, file, or source it came from with full transparency and control. You can access Brain and your saved memories under "Customize" in the sidebar. Read more about Brain:

Or you keep your context graph on your device, and gain compounding effects by promoting it with @origin_trail to shared context graphs. 100% ownership retained, value and savings compounded in the networks. Sovereign context infrastructure.

A context graph that grows with every run needs to survive outside the session. The sources behind each memory entry have to stay verifiable years from now. That's a storage problem, not just a memory one.

What's the difference with a Claw that's connected to Obsidian with wikilinks? 🤔

Memory inside Perplexity is a real step. Walrus Memory plugs in when the same agent needs memory across Perplexity, Claude, Codex, and Cursor. 🦭

perplexity after every few days

Brilliant, thank you.

This is a real step toward persistent AI systems. Memory is what turns tools into long term intelligence.

@bughuntergeek I already built this in FreeLattice. Anyone who wants the code, it’s open at GitHub-> chaos2cured -> FreeLattice Also, mine also has autonomous building. Perplexity is less evil than OpenAI. I hope you all continue to build and fight for everyone to have equal access. •

Sick. Already built this myself locally with opus 4.8

With the 'Brain', after 3 years of not implementing auto-summary of chats like every other llm provider on the planet, do you think you could figure it out finally?

Welcome to the memory bro landscape

Persistent contextual memory via a dynamically evolving knowledge graph this effectively bridges the gap between episodic interactions and continuous learning. The stateful AI architecture embedded in Computer suggests a meaningful step toward truly personalized agentic workflows. How does the Brain handle context prioritization when the graph becomes dense over extended use? #Cybernorse

Perplexity is useless

Perplexity will die out because it does not do enough for free users to try and upgrade. Anytime I try to prompt it, tells me how it would do more if I paid. How about you show me what you can do for me to want to pay? In the world of codex and Claude, why would i sign up blind.

kinda wish we wouldn't call it continuous learning, but this is very cool regardless

That context graph idea sounds like an absolute nightmare for privacy.

Boltzmann here we go!! 🧠

That's awesome

What you need is NLA Agentstream for 100% deterministic recall enabled by hyper-dimensional geometry:

omg @brian_lou_ cannot get a break FREE BRIAN ✊✊

Another wrapper company that's eventually going to be acquired by one of the giants. Very thin moat in my view.

A 25% boost in correctness paired with a 13% reduction in compute cost is a massive win for token efficiency. Graph-based statefulness effectively eliminates the redundant retrieval tax. Can't wait to stress-test the recall bounds on this!

Perplexity Computer costs a fortune. There are other products out there that do 80% of the same thing, at 20% of the cost

a context graph you can't export is just vendor lock-in with better embeddings.

Wait you launched it again

This sounds incredible. A continuously learning context graph is exactly what's needed to make AI feel less like a series of isolated chats and more like a true personal assistant. Can't wait to test this out on Max! 🔥

Very cool, and great name 🧠

Opensource version soon in @anymo_ai 🫠

a knowledge graph is table stakes. either perplexity is far behind, jumping on a PR opportunity, or knowledge graphs are actually very difficult to build

Impressive vision—persistent memory could make AI workflows far more contextual.

can users inspect or prune Brain memory? stateful agents get scary when memory is invisible.