Rethinking Search as Code Generation (research.perplexity.ai)

Perplexity Computer has announced a major innovation in search architecture dubbed "Search as Code" (SaC), which reimagines how AI systems should conduct searches in the age of programmable agents. Traditional search systems have predominantly functioned as monolithic entities that process queries through fixed pipelines, a method that increasingly falls short as AI models and user demands evolve. SaC allows AI agents to dynamically compose tailored retrieval pipelines using atomized search primitives available through a software development kit (SDK), enabling rapid and efficient execution of multiple retrieval operations. This advancement is significant for the AI/ML community as it streamlines the interaction between AI models and search systems, granting models fine-grained control over the search process. Agents can now leverage domain knowledge, optimize search results in real-time, and manage complex workflows that require non-linear operations. By circumventing the limitations of traditional search architectures—such as rigid context management and inefficient control flows—SaC facilitates more sophisticated, contextually relevant results for a wider variety of requests, which could drastically enhance the capabilities of future AI applications.