Show HN: AI-Native Architecture Series (Open-Source) (github.com)

A new open-source repository has been launched detailing the architecture of MyInvestPilot, an AI-native investment system that leverages AI-driven development and a hybrid agent model. This initiative documents the critical design choices, trade-offs, and challenges encountered in creating a production-level system where AI is responsible for generating a significant portion of the implementation code. The repository aims to provide valuable insights for developers rather than serving as a user manual or generic guide, focusing on the practicalities of operating as an AI-native system builder. Key highlights include the use of a constrained Intermediate Representation (JSON DSL), schema-driven contracts, and deterministic execution to ensure reliability in decision-making systems. The system demonstrates an evolution in AI-driven development from a human-centric approach to a more autonomous model, achieving a 90/10 alignment/execution ratio while still maintaining human quality checks. Additionally, the architecture features a split between Local and Remote Agents to manage flexible orchestration and asynchronous processing effectively. This repository is particularly relevant for developers interested in building AI-native applications, agent systems, and cloud-native architectures, emphasizing efficiency in operational costs.