What this case covers
The author Qianjiang does Xiaohongshu virtual materials, previously had no technical background, and first used codex in mid-June. The article reviews her process of using codex plus Obsidian to build a Xiaohongshu writing workbench, the pitfalls she encountered, and skill optimization methods.
Questions it answers
- How do you start building a workbench with zero technical background?
- Why is copying someone else's skill not useful?
- Why does cramming too many tasks cause context collapse?
- Why do search tasks need persistent records?
- How do you judge whether AI output is truly complete?
Our observations
- Copying others' skills fails because business cognition cannot be copied.
- Making implicit judgment explicit is the prerequisite for a reusable workbench.
- Acceptance should look at files and test results, not AI's verbal completion.
From the author
Other experts' skills or workflows are all based on their own work processes, insights, and the cognition and habits cultivated by being immersed in this track; these three are unique.