Perplexity Replaces Fixed Search Service with Programmable Retrieval Primitives

Perplexity has shifted from monolithic search architecture to Search as Code, exposing retrieval components as programmable primitives. Models now assemble custom retrieval workflows on-demand through code generation and sandboxed execution, rather than routing all queries through a single service. The approach scales to thousands of queries per second and supports tasks requiring hundreds or thousands of retrieval operations. This decomposition pattern mirrors how enterprise software evolved from mainframe to microservices—trading optimization simplicity for task-specific flexibility and efficiency.