Perplexity trusts GPT-6 Astra with end-to-end systems

Perplexity describes using OpenAI's GPT-6 Astra to build small testing programs around applications. The model generates substitute responses resembling those an external service would send, allowing the program to exercise a workflow from beginning to end. This mechanism is described in the customer account.

The company says it can check on these tasks less frequently than with earlier models. The useful idea is that an agent can help construct the test environment as well as write application code. The reported increase in trust is the company's experience, without a quantified reliability comparison.