Perplexity has extended Brain, the self-improvement system behind its Computer agent, so it can now automatically create and update agent skills from experience. Skills are reusable instructions that guide the agent through recurring tasks, capturing procedures it would otherwise need to be told again in every session. Some can even include scripts for repetitive work.
The improvement process runs as a four-stage loop. First, Brain gathers evidence from past sessions where a skill was used, especially ones where it fell short. An improvement agent then studies that evidence and proposes an update. Checks confirm the proposal is well formed and likely to help, and the verified version is activated for future sessions. Personalisation changes the agent’s context and instructions, not the underlying model’s weights.
Personal skills are built from a user’s own memory and session history, and the feature is live for US Max subscribers, enabled by default. Creating a skill requires two independent occurrences of the same reusable behaviour, and automatically created skills are capped at ten per user to keep the library curated. Built-in skills, which are available to every Perplexity user, face a higher bar: updates are drawn from aggregate patterns across sessions and must pass paired replay tests plus approval from a human engineer before they take effect.
Perplexity says Brain wrote more than 80,000 personal skills for over 20,000 users between 1 August and 14 November, and applied more than 100,000 automatic updates to over 30,000 existing skills since September. In an eight-week randomised production test, 72.9% of users with the feature enabled received at least one new personal skill, against 6.5% of a control group who added one manually, and tool-call failures fell by 11%. The open question is how well automatically written instructions hold up as workflows grow more complex.
Source: Perplexity