两个社区演示项目展示了 Qwen 模型的实际应用：@richardcsuwandi 基于 Qwen3.8-Max 构建的本地文件助手 Sift，以及 Shadrack 使用 Qwen3.8-27B 与 n8n 开发的 WhatsApp CRM CustomCX。
@richardcsuwandi built Sift with Qwen3.8-Max: a local file copilot that finds files by meaning, reads image content, and proposes changes for approval. This walkthrough shows semantic search, the three-model Ollama stack, and approve/revise/undo in action. Shadrack demos CustomCX, a WhatsApp CRM powered by Qwen3.8-27B + n8n. Follow one conversation from contextual replies to human escalation and AI-drafted responses, then into CRM stages and notes. A practical look at where automation hands control back. @divagr1925 takes Qwen beyond coding: a synthetic web research harness, a new evaluation metric, and a JSON mismatch behind divergent scores for the same model. His walkthrough covers the 18-hour run, trace checks, prompt bias, and where human help mattered.
