Perplexity Brain Makes Agent Memory a Default, Not a Feature

4. Perplexity Brain Makes Agent Memory a Default, Not a Feature

Perplexity CEO Aravind Srinivas announced on June 16, 2026 that Brain is rolling out to all Perplexity users. Brain is a self-improving context graph that aggregates sessions, connected files, and external integrations, then refreshes itself overnight without user prompting. Every subsequent task run through Perplexity Computer draws on that updated context automatically, making the product stateful across sessions rather than resetting on each query.

This is a direct challenge to the memory architecture assumptions baked into competing products. OpenAI's memory layer in ChatGPT requires users to manage what gets saved; Google's Gemini relies on in-session context windows. Brain flips the model: the system decides what is worth retaining and proactively updates it. That overnight refresh cycle means Perplexity Computer accumulates a richer user model over time without friction, which compounds the switching cost for anyone who uses it consistently. For Microsoft Copilot and other enterprise-adjacent agents, Brain signals that persistent memory is moving from a premium differentiator to a baseline expectation in consumer AI.

The broader pattern here is the race to own the "context layer" of personal AI. Whoever holds the most accurate, continuously updated model of a user's work and preferences controls which agent gets called first. Perplexity is betting that a proactive, always-on context graph is more defensible than raw model quality. Watch whether OpenAI accelerates its own memory roadmap in response, and whether Brain's overnight update cycle creates privacy or consent friction that regulators in the EU begin to examine under existing GDPR automated-profiling rules.

Source: @AravSrinivas on X