User Builds Custom Desktop Invoicing App Using Local AI and High-End Hardware
A user with minimal coding experience successfully created a custom desktop invoicing application using local AI models and coding agents. They utilized Qwen 3.8 27B and Open Code v2.0 running via LM Studio on an RTX 5090 GPU. This demonstrates how modern local AI workflows and coding agents enable non-programmers to build practical software entirely offline. It highlights the growing feasibility of replacing paid subscription services with custom AI-generated tools running on high-end consumer hardware. The development leveraged Open Code's auto-compaction feature to maintain a 57K to 90K context window within 31GB of VRAM usage. The user achieved an average processing speed of 80-90 tokens per second using a Q4_K_M GGUF quantization of the Qwen model.
## BACKGROUND
Local LLM tools like LM Studio allow users to run powerful open-source artificial intelligence models directly on their personal computers without sending data to external cloud servers. Open Code is an AI coding agent framework that integrates with local or cloud models to assist with software development tasks directly from a terminal or interface.