The Stagehand team details their toolchain optimizations for browser usage in deep agents. The article emphasizes reducing execution costs by tracing agent behavior, iteratively updating the test harness, and validating with evals, rather than relying solely on model capabilities. This practice offers engineering references for building efficient autonomous web-operating agents.
Evals
评估体系
Standardized processes for quantitatively testing model performance, accuracy, and safety.
Curated Intelligence
Secludy has released the free directory 'Who Sells Data', listing 159 companies selling training data, RL environments, and evals to AI developers, alongside 38 programs that pay businesses for real office records. The directory prioritizes transparency by linking every fact to its source and explicitly marking unverified information rather than guessing.
OpenAI launched AgentKit to address fragmented agent development tools by offering visual workflow building, connector management, and embedded UI components. The suite integrates Evals and Reinforcement Fine-Tuning (RFT) under standard API pricing, with plans for a standalone Workflows API. Notably, OpenAI announced the deprecation of legacy Agent Builder and Evals products starting late 2026, directing users toward the new architecture or ChatGPT Workspace Agents.
OpenAI has announced the Pioneers Program, designed to collaborate with startups in high-impact verticals like legal, finance, and healthcare to create domain-specific evaluation benchmarks and optimize reasoning model fine-tuning. The initiative aims to establish standards for real-world AI performance and provide tools for builders to enhance model capabilities in specific domains. It is currently in an early recruitment phase targeting companies with clear applied use cases.