Perplexity trusts GPT-6 Astra with systems: writes comms, changes software and monitors production

Perplexity uses GPT‑6 Astra to perform tasks beyond text generation: the model writes communications, changes software and monitors production systems.

Johnny Ho, Perplexity cofounder and Chief Strategy Officer, says that improvements in the model’s coding ability also improve the company’s search: the model can write programs that search the web and internal information and summarize results concisely. Ho links better code-writing to better search.

Ho identifies the real challenge as applying those informational capabilities to real-world systems, and he says GPT‑6 Astra helps with that. Perplexity asks the model to build small testing programs around applications; the model generates realistic responses from other services — for example a language model API or a connector — to check how an application reacts and to test end-to-end workflows.

Perplexity also reports it checks GPT‑6 Astra far less frequently than earlier models, though no exact checking intervals or methods are provided in the source.

Together these uses let Perplexity move the model’s capacity to process large amounts of information into practical operations: communications, software changes and production monitoring, according to Johnny Ho.