Perplexity Computer Brain Adds Self-Writing Agent Skills With 72.9% Adoption
Perplexity Computer's Brain now writes and updates agent skills for users based on past sessions, with an improvement agent analyzing full session trajectories and proposing changes only when evidence is sufficient. In an 8-week randomized production test, 72.9% of users received at least one new personal skill versus 6.5% who added one manually, while tool call failures fell 11% and dissatisfied threads fell 7.6% with no measurable change in cost or latency.
Event version 1.
Key insights
Brain writes personal skills from session history
Perplexity Computer's Brain, which already maintains a knowledge wiki about user work, now writes and updates agent skills based on past sessions. An improvement agent reads full session trajectories including tool calls, reasoning, and results, and proposes a new skill or update only when there is enough evidence.
Built-in skills undergo replay and engineer review
Built-in skills that every user receives go through paired replay on failure scenarios and engineer review before any change goes live.
Randomized test shows high skill adoption
In an 8-week randomized test on production traffic, 72.9% of users got at least one new personal skill, compared with 6.5% who added one manually.
Failures and dissatisfaction decline
Tool call failures dropped 11% and dissatisfied threads dropped 7.6%, with no measurable change in cost or latency.