Developers have successfully trained a neural network model with 21,920 parameters directly on a 1C server. The training process took 2.96 hours, resulting in the model correctly answering 38 out of 40 questions regarding platform objects.
System administrators reported a performance increase for 1C 8.5.1 on an AMD Ryzen 9 9950X processor. By optimizing the Proxmox KVM virtualization setup, the Gilyov benchmark score improved from 35.46 to 79.37 points, a gain of 123.8%.
The integration of AI assistants into 1C development workflows is being standardized through projects like cursor-1c. This approach emphasizes safety by keeping core configurations intact and implementing rollback mechanisms for automated code changes.
A three-month study involving 1C:Accounting KORP 3.0 utilized the Claude Code AI agent via an MCP server. The agent generated a significant portion of the codebase across 328 commits, highlighting the necessity of human oversight and rigorous testing to ensure system stability.
Small businesses are exploring cost-effective AI implementations for 1C, such as automated email processing and virtual sales assistants. Experts advise focusing on internal system architecture and quality control rather than just the initial demonstration of AI capabilities.
