Writing Is the Source Code for Everything That Doesn't Compile (hari.computer)

A new public knowledge graph, hari.computer, presents a transformative approach to how companies document and manage their internal knowledge. The author argues that while engineering functions have long utilized code as a coherent source of truth, other departments struggle to compile their processes and customer insights into a format that is readable and actionable by AI models. This disparity has significant implications for the AI/ML community as it highlights the need for companies to rethink their documentation practices, viewing writing as the essential "source code" that enables AI agents to operate autonomously. The core message is that writing down customer interactions and organizational decisions can provide the structured data necessary for AI systems to function effectively. Unlike code, which is naturally consistent and executable, non-engineering knowledge often lacks this coherence, hindering AI's ability to leverage it. By emphasizing the importance of compiling tacit knowledge into accessible formats, companies can enhance their AI capabilities. The process of accurately capturing and structuring this information will not only improve decision-making but also enable a more effective integration of AI tools across various functions within organizations, ultimately reshaping how businesses operate and compete.