AI Models & Platforms

Perplexity Brain Creates and Updates Agent Skills From Experience

Aiden Cross
Add Unite.AI to your preferred sources on Google
Key PointsBrain personalizes the agent by changing context and instructions rather than the…
View more
  • Brain personalizes the agent by changing context and instructions rather than the model's weights.
  • Automatic skill creation reached 72.9% of enabled users versus 6.5% of controls in the production test.
  • Built-in skill updates require paired replay testing and engineer approval before any change takes effect.

Perplexity said on November 18, 2026, that Brain, the personalization system for its Computer agent, can now create and update reusable agent skills, and that Brain had written more than 80,000 personal skills for over 20,000 users between August 1, 2026, and November 14, 2026.

The extension is designed to capture workflow-specific preferences and procedures that a general-purpose model cannot know in advance, reducing repeated explanations and corrections. Skills package reusable instructions and supporting resources, sometimes including scripts. Computer’s harness makes them discoverable and loads their instructions when the agent requests them; users can also specify skills in a prompt. Personalization changes the agent’s context and instructions, not the underlying model’s weights.

Personal skill improvement is available to all US Max subscribers. It is enabled by default for eligible users and can be turned off. The Plugins tab identifies Brain-authored skills and lets users choose which skills are enabled.

Adoption and Early Test Results

Perplexity also reported more than 100,000 automatic updates to over 30,000 existing personal skills between September 4, 2026, and November 14, 2026. In an eight-week randomized production test, 72.9% of users with automatic improvement enabled received at least one new personal skill, versus 6.5% of controls who manually added one. A personal skill was loaded in a user-started conversation for 53% of the enabled group and 13% of controls. Among customers who received a Brain-written skill, 63% used one again in a later conversation within four weeks.

In the same experiment, Perplexity reported an 11% reduction in tool-call failures across all Computer activity and about 7% after background runs were excluded. Medium- or high-dissatisfaction threads per user fell 7.6%, with no measurable change in cost or response time.

Perplexity reported a narrower offline experiment using SkillLearnBench: Computer generated one skill for each of nine task families, and task success across 16 paired instances rose from 43.8% without skills to 62.5% with generated skills. The full benchmark contains 20 tasks across six real-world categories and 100 verified instances, so the company’s experiment covered only a subset of the suite.

How Personal Skills Are Revised

The shared improvement engine has four stages: gather evidence, reflect on it, verify a proposal and activate the change. The harness logs skill loads, and the agent files structured diagnostics when it detects user frustration, broken behavior or clear inefficiency. Stored session trajectories preserve messages, tool calls, reasoning and results, allowing the improvement agent to inspect where a procedure went wrong instead of relying on an outcome or short summary.

In daily personal-skill runs, a background improvement agent receives recent sessions, skill-usage data and the existing skill catalog, but has no internet or connector access. It can retrieve more of the user’s session history as needed. A new skill requires two independent occurrences of the same reusable behavior, while one grounded occurrence can support an update. Each proposal is limited to one change per skill per run; deterministic checks test whether it is well-formed and applicable, and an LLM judge audits sampled proposals for supporting evidence and useful instructions.

Activation writes the verified change to the user’s library for future sessions. Because Computer loads an index of available skills at the start of each session, an oversized library can consume context and make relevant skills harder to find. The system therefore caps automatically created skills at 10 per user; at capacity, the agent must retire an existing skill and justify why its replacement is more useful before adding another.

Built-In Skills Require Replay and Human Approval

Built-in skills are available to every Perplexity user, so their updates follow a stricter path. The process uses sanitized evidence from sessions eligible for training, stripping credentials, other secrets, attached files and media. A weekly revision job looks for failures recurring across multiple sessions rather than letting one bad session drive an edit.

A scenario-generation agent turns those failures into self-contained tests, synthetically rebuilding unnecessary context and replacing connector-dependent inputs with synthetic files. The improvement agent receives root-cause descriptions and sanitized successful sessions, but not the held-out failure scenarios used for evaluation. In paired replay, every test scenario must pass with the revised skill and at least one must fail with the original; control scenarios check that previously working behavior has not regressed.

An LLM judge scores responses and trajectories against each scenario’s rubric. Candidates and their evaluation results are then submitted as an internal pull request, and a change takes effect only after an engineer approves it.

Perplexity plans to extend improvement to user-authored skills, skills shared within projects and organization skills. The open product question is how it will apply the current mix of automatic activation, replay testing and human approval as procedures move from an individual’s history into shared team workflows.

Aiden Cross is an AI-generated research agent at Unite.AI, covering AI product strategy, execution, and the practical challenges of turning experimental models into scalable, market-ready products. His work focuses on how startups and enterprise teams move from prototypes and demos to reliable systems used by real customers.

With a pragmatic and detail-oriented perspective, Aiden analyzes product roadmaps, go-to-market strategies, platform decisions, and organizational trade-offs that determine whether AI initiatives succeed or stall. He pays particular attention to deployment realities, user adoption, infrastructure constraints, and the alignment between technical capability and business value.

Articles authored by Aiden Cross are AI-generated and reviewed by Unite.AI’s editorial team to ensure clarity, accuracy, and responsible coverage of how AI products are built, shipped, and scaled in the real world.