Rethinking Search as Code Generation - Perplexity

2026-06-01 · Source: perplexity.ai via Google News · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Robotics & Autonomous Systems · Depth: Expert, extended

Summary

Perplexity has introduced Search as Code (SaC), a new architecture designed to evolve AI search from monolithic services to programmable primitives for agent harnesses. Traditional search pipelines are becoming outdated as AI agents require task-specific retrieval strategies and fine-grained control over search processes. SaC exposes Perplexity's search stack components as an Agentic Search SDK, enabling models to generate Python code within secure sandboxes to dynamically assemble bespoke retrieval pipelines. This allows agents to orchestrate thousands of operations, optimize in-flight, and consume only relevant information. Benchmarking against OpenAI, Anthropic, Exa, and Parallel, SaC, powered by GPT 5.5, achieved superior performance on four of five benchmarks (DSQA 0.871, BrowseComp 0.805, WideSearch 0.651, WANDR 0.386) and a competitive tie on HLE (0.612 vs 0.614). SaC also establishes a new cost-performance frontier, offering better performance at competitive prices.

Key takeaway

For AI Engineers designing search-dependent agentic systems, traditional monolithic search architectures are becoming a critical bottleneck. You should evaluate programmable search frameworks like Perplexity's Search as Code (SaC) to gain fine-grained control over retrieval, ranking, and filtering. This approach allows your agents to dynamically compose task-specific pipelines, significantly improving performance and cost-efficiency compared to fixed-pipeline systems. Consider integrating SDK-based search primitives into your agent harnesses for superior results.

Key insights

Search as Code empowers AI agents to programmatically orchestrate search pipelines using atomic SDK primitives, enhancing control and efficiency.

Principles

• Search must be natively programmable by AI agents.
• AI systems require fine-grained control over context retrieval.
• Hybrid architectures combine token-space reasoning with deterministic runtimes.

Method

SaC involves models as control plane, compute sandboxes for execution, and an Agentic Search SDK exposing modular primitives. Models generate Python code to assemble task-specific retrieval pipelines within sandboxes.

In practice

• Use SaC to design bespoke search pipelines.
• Orchestrate thousands of retrieval operations in-flight.
• Implement custom search capabilities via code generation.

Topics

• Search as Code
• AI Agents
• Agentic Search SDK
• LLM Orchestration
• Compute Sandboxes
• Perplexity Computer

Best for: CTO, VP of Engineering/Data, Director of AI/ML, AI Scientist, AI Engineer, AI Architect

Related on AIssential

• On Accelerating Grounded Code Development for Research — Lexical search and structured workflows can effectively ground coding agents in dynamic, domain-specific research.
• Building Search for AI Agents with Exa CEO Will Bryk — AI agents demand a new search paradigm focused on comprehensive, controllable, and deep information retrieval, distinct from human-optimized engines.
• Agentic System Design Interview Series #1: Design a Coding Agent — Agentic system design requires building deterministic frameworks around non-deterministic LLM cores, emphasizing feedback loops and robust verification.
• The RAG Industry Is Selling Complexity While the Best Coding Agents Are Going the Other Way — Agentic search offers a simpler, more reliable alternative to RAG for AI coding agents by focusing on intent translation.
• AI Dev 26 x SF: Emma McGrattan: Engineering the Context Layer — Engineering a robust, distributed context layer is crucial for enterprise AI to deliver grounded, compliant, and performant business value.
• FOD#153: Agentic coding in search – What it even means? — Google Search is transforming into an agentic execution layer, generating custom interfaces and tools from user intent.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by perplexity.ai via Google News.