Show HN: Open-Source SDK for AI Knowledge Work (github.com)

A new open-source Python SDK has been introduced for building AI agents capable of performing knowledge work, including research, analysis, writing, and decision-making tasks. This SDK addresses the limitations of traditional automated testing in knowledge work, which lacks clear correct answers and verification methods. By employing structured rubrics that define criteria for success prior to execution, the SDK allows agents to self-verify their work, iteratively improve based on failures, and provide a transparent evaluation process that can be audited by users. The implementation integrates various capabilities, such as web searches and file handling, all managed by an orchestrator that ensures output verification. This initiative is significant for the AI/ML community as it focuses on enhancing AI's role in knowledge-intensive professions, an area that has been relatively underserved compared to traditional coding tasks. The SDK provides a critical building block for tools that automate research and recommendation processes, promising substantial time savings in product development and deployment. Additionally, with the self-verification mechanism, there’s potential for training models on rubric-based verification, potentially elevating the overall robustness of AI applications in knowledge management and beyond.