The developer initiated a project to create a remote access tool using AI, aiming to test if artificial intelligence could independently build a complex software architecture. The process began in early summer, with the developer delegating coding, architecture, and technology selection to AI models. The initial attempt resulted in a project with over 120,000 lines of code that failed to function correctly due to recurring technical errors.
Following the initial failure, the developer shifted the approach to manual architectural design while using AI to handle the mechanical aspects of coding. This transition allowed for faster refactoring and testing of different network strategies. The development process focused heavily on the data transmission protocol, with the creator ultimately choosing UDP over WebSockets to better navigate network obstacles.
The final architecture relies on a signaling server for session coordination, while actual screen data is transmitted via direct client-to-client connections or server relays. To improve global stability, the developer relocated the server infrastructure from a Moscow data center to the UAE. The system is designed with horizontal scaling capabilities, theoretically supporting up to one million concurrent users.
The project serves as a case study in AI-assisted development, highlighting the necessity of human oversight in complex system design. The developer noted that while AI significantly accelerates code modification, it often introduces new issues that require deep manual analysis. This experience provided practical insights into managing network traffic, relay management, and infrastructure scalability.
