Perplexity launches 'Search as Code' for agents

// 11d agoPRODUCT LAUNCH

Perplexity launches 'Search as Code' for agents

Perplexity has introduced 'Search as Code' (SaC), a programmable search architecture that allows AI agents to write and execute Python code using low-level search primitives. By replacing multi-turn tool calling with sandboxed execution, the new system reduces latency and token consumption for complex retrieval tasks.

// ANALYSIS

Hot Take: Search is no longer a static endpoint; it is a programmable runtime library. By shifting to Search as Code, Perplexity aims to own the developer framework layer for AI agents, transforming search APIs from black-box document retrievers into sandbox-executable SDKs.

–Sandboxed execution of search primitives enables massive efficiency improvements for complex multi-step reasoning.
–Condensation of search loops into single-run generated scripts drastically reduces token consumption and network round trips.
–Elevates agent resilience by letting LLMs dynamically write, test, and debug their own custom data aggregation algorithms.

// TAGS

perplexityagentsearch-as-codecode-generationdevtoolagentic-workflowsai-coding

DISCOVERED

11d ago

2026-06-01

PUBLISHED

11d ago

2026-06-01

RELEVANCE

9/ 10

AUTHOR

AravSrinivas

// KEEP READING

More AI developer news from the feed

EXPLORE FULL FEED

OPEN SOURCE48m ago

Music Assistant is an open-source media library manager that unifies your music streaming services and smart speakers into a single self-hosted server.

Music Assistant functions as a centralized hub for your music library, connecting various streaming services and local media directories to a wide range of smart speakers. Designed to run continuously on low-power devices like a Raspberry Pi, NAS, or Intel NUC, the server aggregates searches, manages playlists, and orchestrates multi-room synchronized audio playback across disparate hardware systems.

NEWS51m ago

Developers compare Claude Fable 5 and ChatGPT 5.5

A social media inquiry by @droidbuilds asks the developer community to identify distinct capabilities that Anthropic's new "Mythos-class" Claude Fable 5 model can perform which OpenAI's ChatGPT 5.5 cannot. This query highlights the intense competition and performance comparisons between the latest frontier AI models as developers benchmark their long-horizon agentic workflows, complex coding tasks, and multi-step execution.

POLICY1h ago

Amazon reports trigger Fable 5, Mythos 5 ban

Following reports that security researchers successfully jailbroke Anthropic's restricted Mythos 5 AI model, the U.S. Department of Commerce issued an emergency export-control directive restricting access for foreign nationals. Because Anthropic was unable to dynamically filter users by nationality, it requested its cloud partner, Amazon Web Services (AWS), to disable global access to both Fable 5 and Mythos 5, sparking debate over government oversight of frontier AI models.