Show HN: I don't code – I run a self-correcting protocol across 4 projects (github.com)

A new file-based execution protocol has been implemented across four independent projects—including a crypto trading system and an AI decision system—without any coding by its creator. Instead, the protocol serves as an innovative decision-making and oversight framework, allowing an AI to autonomously write and execute code. This approach contrasts traditional enterprise AI governance models, which are often cumbersome and effective mainly at larger scales. The described system emphasizes continuous self-correction; failures in one project lead to alterations in the rule hierarchy that can be applicable across diverse domains. The significance of this development lies in its potential to redefine agentic AI governance by enabling dynamic learning and adaptability across projects. Unlike existing models that restrict learning to a single environment, this protocol facilitates cross-project insights that enhance system reliability without the overhead typically associated with enterprise solutions. The governance framework is structured around three key files—CORE.md, AGENT.md, and SESSION_INDEX.md—that together ensure a cumulative, evolving process, pushing forward the boundary of how AI systems can learn from past experiences and adapt accordingly. This unique approach may pave the way for more efficient and responsive AI governance practices, particularly for smaller-scale implementations.