What this case covers
The author Qiantangjiang Li summarizes nearly half a year of AI programming and content creation practice, covering data collection, account matrix operations, enterprise agent delivery, and the content publishing pipeline. The article gives tool division of labor, delivery unit prices, and efficiency changes.
Questions it answers
- How do Claude Code and Codex divide labor?
- How can a data collection system be built from scratch?
- How does the thousand-account matrix Agent operate?
- Why did enterprise agents shift to native implementation?
- How can the content creation pipeline be shortened?
Our observations
- Tool division of labor is stronger than a single model; pairing fast and slow covers most problems.
- The shift from low-code to native implementation reflects that the B-side values stability more.
- The increase in parallel projects shows that the bottleneck has shifted from coding to requirements and architecture.
From the author
If you treat it as a tool, it can only write some code. If you treat it as an employee, it can help you do an entire process.