Perplexity AI Updates: June 8, 2026

1. Perplexity Launches Search as Code to Cut Agentic Token Costs

Perplexity introduced Search as Code, an architecture in which AI agents write custom Python search pipelines from composable primitives such as retrieve, filter, deduplicate, and rerank, rather than calling fixed APIs. In a CVE research benchmark the approach reduced token usage by 85 percent versus Perplexity’s standard pipeline, and the company reports accuracy gains of 7 to 45 percent across five other benchmarks. The feature is rolling out in Perplexity Computer and the Agent API, giving developers a way to tailor retrieval logic to specific tasks instead of accepting a one-size-fits-all search step. Source