Perplexity Automates Project Management with GPT-6 Astra

Perplexity entrusts GPT-6 Astra with writing communications, software modifications, and production oversight. Co-founder Johnny Ho says he conducts fewer checks than with previous models and relies on Astra for end-to-end testing.

Automated Testing: Astra Simulates Services and Covers the Full Workflow

When manual testing time is limited, Johnny Ho explains that he asks GPT-6 Astra to generate a test program around an application. The model produces realistic responses, similar to those that a third-party service like a language model API or a connector would send. By replacing these services, Astra can observe the application's reaction and test the entire workflow. Johnny Ho claims he can entrust Astra with complete end-to-end systems and reduce the frequency of checks compared to previous models. Perplexity also indicates that it conducts overall checks much less frequently with Astra than with earlier generations.

Operations: Communications, Code, and Oversight Entrusted to Astra

Astra is employed by Perplexity for writing communications, modifying software, and monitoring its production environments. Johnny Ho also notes that the model is used to make changes to real systems and ensure the follow-up of operational software. According to him, the main challenge is applying these informational capabilities to concrete systems, and GPT-6 Astra makes this step more accessible.

Research: Expected Gains as Code Writing Advances

Perplexity presents itself as an AI-based response engine focused on research and accuracy, where managing large volumes of information plays a key role. Johnny Ho, co-founder and Chief Strategy Officer, emphasizes that every advancement of the model in code generation also leads to progress for Perplexity's search engine. He mentions the possibility of creating more efficient programs capable of querying the web as well as internal data, and then generating particularly concise summaries.