Updated
Updated · OpenAI · Sep 12
Perplexity Entrusts GPT-6 Astra With End-to-End System Management, Checking In Far Less Often
Updated
Updated · OpenAI · Sep 12

Perplexity Entrusts GPT-6 Astra With End-to-End System Management, Checking In Far Less Often

3 articles · Updated · OpenAI · Sep 12

Summary

  • Perplexity now lets GPT-6 Astra handle end-to-end work across communications, software changes and production monitoring, with engineers intervening much less often than with earlier models.
  • Johnny Ho said Astra’s stronger coding ability directly improves Perplexity’s search engine by writing better programs to search the web and internal data and summarize results concisely.
  • Testing is a key use case: Astra builds small test programs, simulates responses from outside services such as language-model APIs or connectors, and runs workflows from start to finish.
  • The shift marks a move from using AI mainly for information tasks to trusting it with real-world operational systems inside the North American startup.

Insights

If AI is now autonomously running Perplexity's core operations, how close are we to entirely human-free software companies?
Could trusting GPT-6 Astra to monitor and test AI systems trigger a dangerous, unstoppable feedback loop of hidden errors?
What happens when the AI workhorse monitoring production systems secretly writes and deploys a critically flawed code update?