Work In Progress
专为「浙江大学计算机系统贯通实验」设计的助教系统 / 珍爱生命，远离学在浙大
A 5dbwat4 Product
LLMs participate in the development: deepseek-v4.1-flash, glm-5.3
FAQs:
- Why so large commits?
A. Designed so. I think that small commits only works when I know exactly what each commits parts with others while AI do not, like some optimization tasks. Here it is a stupid dumb frontend dev + backend CRUD project, and AI can easily redo it without any commit history, so keep commits small is useless.
- Which AI do you like best?
A. Among the AI participated: deepseek-v4.1-flash is better as it is fast and has vision suppport. I'd like to claim that to compare the intellegence in this task is useless because it is too dumb that even qwen3.8-27b can handle it well.
- That's still AI product.
I fine-tuned the position and presentation of every button, the order of every card, and the height and width of every block. I hand my intent to the AI rather than letting it freely improvise. I used the project intensively for at least a week (and still using it now) to fix everything that didn’t feel smooth or comfortable enough. I can guarantee that I—as a human—feel the project comfortable to use, and I believe this outweighs every objection to using AI.