A developer tests open-source LLMs like Qwen3 for terminal-based coding workflows, revealing practical strengths in troubleshooting but limitations in autonomous development. The experiment highlights the viability—and current constraints—of running local AI coding assistants with tools like Ollama and aider.