AI Product Builder — agent systems, desktop tools, and practical automation.
I turn ambitious ideas into focused products that are simple to use, reliable in practice, and verified before they are called done.
我是一名仍在持续学习和交付作品的学生，也是一名 AI 产品构建者。我专注于把复杂工作流做成真正可用的 Agent 系统、桌面工具和自动化产品。
| Project | What it delivers |
|---|---|
| project-autopilot | Gives complex Codex projects a project lead, bounded coordination, durable context, permission gates, and acceptance evidence. |
| task-effort-router | Keeps small tasks light and gives large tasks the planning, reasoning, goals, and verification they actually need. |
| SparkKeeper | A Windows GUI for calibrated, scheduled Douyin and WeChat streak automation with local statistics. |
-
AI agents and skills — task routing, project coordination, context continuity, and evidence-driven completion.
-
Desktop products — focused interfaces that make useful automation approachable.
-
Practical automation — local-first tools built around real workflows instead of demos.
-
AI Agent 与 Skill：任务分级、项目协调、上下文延续和可验证交付。
-
桌面产品：用清晰界面降低自动化工具的使用门槛。
-
实用自动化：围绕真实工作流构建，而不是只做概念演示。
- Start with the real outcome, not ceremony.
- Keep the interface simple and the execution dependable.
- Treat verification as part of the product.
先确认真正要交付的结果，再选择合适的投入；界面保持简单，执行必须可靠；没有经过验证，就不算完成。
- Portfolio: Dylan Cai
- Project questions and collaboration: open an Issue in the relevant repository.
If one of these projects solves a problem you care about, try it, star it, or tell me where it falls short. Useful feedback shapes what I build next.
如果这些项目正好解决了你的问题，欢迎试用、点 Star，或者通过 Issue 告诉我哪里还不够好。真实反馈会直接影响下一步改进。