> ## Content Index
> Fetch the complete content index at: https://www.tao.media/llms.txt
> Use this file to discover other available public pages before exploring further.

# Perplexity Brain Now Builds and Updates Personal Skills for Computer
- URL: https://www.tao.media/perplexity-brain-now-builds-and-updates-personal-skills-for-computer/
- Published: 2026-11-20T21:33:05.000Z
- Updated: 2026-11-20T21:33:05.000Z
- Description: The update turns past sessions into reusable workflow instructions, reducing the need for users to configure skills themselves.
- Author: Bart Hillerich
- Tags: Perplexity, AI, News

Perplexity has added [automatic personal-skill creation and updates](https://x.com/perplexity%5Fai/status/2123449395882701224?ref=tao.media) to Brain, the personalized self-improvement agent in Computer. The system draws on past sessions and memory to turn recurring ways of working into instructions Computer can reuse on future tasks.

The update reduces the need for users to configure those instructions themselves. In an eight-week randomized test on production traffic, Perplexity found that 72.9% of users with automatic skill configuration enabled gained at least one new personal skill, compared with 6.5% of users in a control group who added one manually.

![Personal skill improvement loop](https://storage.ghost.io/c/78/0b/780ba906-b1a7-4bf0-873c-bdd5c32e5331/content/images/2026/11/HXgB910a4AAzs66.png)

[Personal skill improvement loop | Perplexity (X)](https://x.com/perplexity%5Fai/status/2123449395882701224?ref=tao.media)

## From Personal Memory to Reusable Procedures

A skill is a reusable set of instructions for a recurring task. It captures workflow preferences that a general-purpose model would not know on its own, giving Computer a procedure to follow rather than requiring the user to explain it again.

Brain already maintains a knowledge wiki about a user's work. [Personal skills extend that memory into procedures](https://x.com/perplexity%5Fai/status/2123449419995685295?ref=tao.media). Besides storing information about the user, the system can store instructions for how to do that person's recurring work.

![Skills and Brain procedural memory](https://storage.ghost.io/c/78/0b/780ba906-b1a7-4bf0-873c-bdd5c32e5331/content/images/2026/11/HXgB_mhagAAOfsS.png)

[Skills and Brain procedural memory | Perplexity (X)](https://x.com/perplexity%5Fai/status/2123449419995685295?ref=tao.media)

Brain's improvement process has four stages. It gathers evidence from past sessions, uses an improvement agent to propose a change, checks the proposal, and applies the verified revision.

Future tasks can then use the updated instructions.

![Four-stage loop evidence, proposal, checks, activation](https://storage.ghost.io/c/78/0b/780ba906-b1a7-4bf0-873c-bdd5c32e5331/content/images/2026/11/HXgCAmkbMAA2TXw.png)

[Four-stage loop evidence, proposal, checks, activation | Perplexity (X)](https://x.com/perplexity%5Fai/status/2123449444163281194?ref=tao.media)

The improvement agent examines the full session record, instead of relying only on a final result or a shortened summary. That gives it the interaction history behind a task when deciding which recurring behavior is worth turning into a personal skill. When the evidence does not support a change, it leaves the skills unchanged.

![Improvement agent reads full session record](https://storage.ghost.io/c/78/0b/780ba906-b1a7-4bf0-873c-bdd5c32e5331/content/images/2026/11/HXgCCWobUAA8Od0.png)

[Improvement agent reads full session record | Perplexity (X)](https://x.com/perplexity%5Fai/status/2123449467244515815?ref=tao.media)

## Different Checks for Personal and Shared Skills

Perplexity uses different verification paths depending on how widely a proposed change will apply. A personal skill affects one user's workflow; a built-in skill is available across the product.

Personal skills rely on deterministic, or rule-based, checks to validate proposals, while an LLM judge audits a sample.

The requirements for a built-in revision include both testing and human approval. The revised skill must pass every test scenario, while the original must fail at least one. An engineer must then approve the change before it takes effect.

![Verification paths for built-in and personal skills](https://storage.ghost.io/c/78/0b/780ba906-b1a7-4bf0-873c-bdd5c32e5331/content/images/2026/11/HXgCDW3bUAAJV1q.png)

[Verification paths for built-in and personal skills | Perplexity (X)](https://x.com/perplexity%5Fai/status/2123449484718075950?ref=tao.media)

## What Changed in the Eight-Week Test

The experiment measured whether skills were used as well as whether they were created. In Perplexity's enabled group, 53% of users had a personal skill loaded in a conversation they started, compared with 13% in the disabled group. Loading makes a skill's instructions available for the task.

![Automatic skill improvement creation and loading chart](https://storage.ghost.io/c/78/0b/780ba906-b1a7-4bf0-873c-bdd5c32e5331/content/images/2026/11/HXgCEXbagAAXyE4.png)

[Automatic skill improvement creation and loading chart | Perplexity (X)](https://x.com/perplexity%5Fai/status/2123449513931333734?ref=tao.media)

Perplexity reported an [11% reduction in the tool-call failure rate](https://x.com/perplexity%5Fai/status/2123449529999712548?ref=tao.media) across all Computer activity in the same test. Excluding the system's own background runs, the reduction was about 7%. Tool calls are the operations an agent uses to act through its available tools.

Conversations with medium or high dissatisfaction fell by 7.6% per user. Perplexity found no measurable change in cost or response time during the experiment.

Personal skill improvement is [available to all consumer and enterprise subscribers in the U.S.](https://x.com/perplexity%5Fai/status/2123449541651534007?ref=tao.media). It is enabled by default for eligible users and can be switched off in settings.