Perplexity uses GPT-6 Astra to write communications, edit software, and monitor production systems with minimal human oversight. As an AI-powered answer engine focused on search accuracy, Perplexity benefits from improved code generation that enhances its web and internal information summarization. Cofounder and Chief Strategy Officer Johnny Ho states that each model advancement in writing code directly improves the search engine’s ability to create better programs for information processing.

The real challenge lies in applying these capabilities to real-world systems, which GPT-6 Astra simplifies. Ho explains that the model can now craft communications, edit real-world systems, and monitor production software in ways prior generations could not. A key application is automated testing: GPT-6 Astra generates realistic API responses to simulate services like language model APIs or connectors, enabling end-to-end workflow validation.

Perplexity can now trust the model with full system testing and check in far less frequently than with earlier models, significantly increasing operational efficiency.