Automating website creation from WhatsApp audio using Qwen3.6 35B and pi
Leverage Qwen3.6 35B and pi to automate end‑to‑end website creation from audio transcripts.
Experiment with pi and Qwen3.6 to automate repetitive tasks and build websites from audio.
Summary
A user describes a workflow that uses Codex to create skills for devops, docling, playwright, and code tickets, which are then executed by Qwen3.6 35B. The user transcribes WhatsApp audio into a markdown content file, then instructs pi to generate a landing page from that content. A plan.md file lists tickets, and the main pi orchestrates sub‑pi instances to execute each ticket, marking them DONE and committing changes. All tasks, including uploading the site to a VPS, are performed locally on Qwen3.6 35B. The user praises pi’s meta‑software capabilities, noting its ability to extend itself and manage skills. The workflow demonstrates end‑to‑end automation from audio to live website without external cloud services.
The post showcases how local large models can drive complex, multi‑step workflows with minimal human intervention.
Key changes
- Codex creates skills for devops, docling, playwright, code tickets
- Qwen3.6 handles natural language OS commands
- Audio transcripts used to generate landing page content
- plan.md tickets orchestrated by main pi to spawn sub‑pi instances
- All tasks performed locally on Qwen3.6 35B
- pi software can extend itself and manage skills
- Future Qwen3.7 expected to improve performance