The latest articles on DEV Community by Nothing To Do (@0c33). https://dev.to/0c33 https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3938471%2F001811cb-73b2-4e01-bbf5-55936ce6bda2.png https://dev.to/0c33 en Nothing To Do Mon, 18 May 2026 15:34:56 +0000 https://dev.to/0c33/i-built-an-agent-that-builds-agents-runs-on-local-qwen3-pure-python-2hjp https://dev.to/0c33/i-built-an-agent-that-builds-agents-runs-on-local-qwen3-pure-python-2hjp <p><a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fi8ihrg2wdhfal0odmlzt.png" class="article-body-image-wrapper"><img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fi8ihrg2wdhfal0odmlzt.png" alt=" " width="800" height="1200"></a>Hi, I built this agentic AI system.</p> <p>Closed-loop — takes a raw idea, ships a standalone .py agent file.</p> <p>What's different:</p> <ul> <li>Interviews you until it understands the request before building anything</li> <li>Two testing stages: prompt validation via LLM invoke, then real subprocess execution of generated code. Not the same thing.</li> <li>Self-referential: injects its own source as a reference template for generation</li> <li>Structured rating schema drives iteration. Human approval gate before anything saves.</li> </ul> <p>Runs on Qwen3.6-35B A3B Q8_0 locally via llama.cpp.</p> <p><a href="https://github.com/0c33/Agentic-Ai" rel="noopener noreferrer">GITHUB PAGE</a></p> <p>Give it a shot and tell me what you think.</p> agents ai python showdev