[Review] AutoBE Hackathon 2025-09-12 - Simple chat bot backend

[mjy9088]

Conversation Links

https://hackathon.autobe.dev/?session-id=0199473c-c4c4-72fa-adb1-52cd5a788964

1. Requirements Analysis

• It seems to have roughly understood the requirements.
• It keeps introducing roles unnecessarily (which are not needed for my requirements).
• It appears unable to catch important missing parts in the requirements (such as features that have no way to be invoked).

2. Database Design

• The relationships seem fine.
• It adds fields that are unnecessary or even impossible to implement given the requirements (e.g., adding deleted_at even though users have no way to know if something was deleted).
• Even when given feedback on the table structure, it does not reflect the changes in the actual PRISMA step output, despite responding as if it would (even after explicitly restarting the PRISMA step).

3. API Design

• The documentation and formats seem fine.
• It followed RESTful principles even in places where it should not have.
• It also implemented typical but unnecessary features (like authentication) that were not part of the requirements.

4. Test Code

• It was a simple app without much business logic, but it seemed to test most of the functionality.
• It appeared to cover most error cases as well.
• There seemed to be parts where the API was modified just to make random-dependent logic testable without using mocks.

5. Implementation Code

• The readability and structure seem fine.
• Security also appears to be sufficiently considered.
• I didn’t read all the code, but I didn’t notice any issues like N+1 problems.

Estimated Code Proficiency Level

junior

🙌 Overall Review

• Strengths: When there’s a large amount of work to do, it can significantly reduce the amount of typing needed while still producing a reasonably acceptable level of code quality—especially for typical, standard projects.
• Weaknesses: (100% subjective, and may only apply to this specific case) It didn’t seem very flexible. Possibly because my use case was not typical, it kept adding unnecessary features and leaving out important ones.
• Code Quality: Honestly, my idea this time was very poor, so there wasn’t really any part where I could evaluate the quality of critical code. I didn’t see anything that was clearly badly written, but also nothing that stood out as well-designed (though the idea itself didn’t really allow for that).
• Improvements:
  • When I gave feedback and said continue, it sometimes moved on to the next step instead of redoing the previous one with my changes. There’s currently no way to stop an ongoing process, and adding such a feature could help avoid wasting tokens.
  • I had more feedback, but can’t recall them right now. I’ll add them later if I remember.
• Overall: It’s still far from being able to work proactively on its own, but as a tool to reduce my workload by a certain percentage, it seems more capable than a complete beginner.

Thanks for the opportunity to try AutoBE. Unfortunately, I was very busy lately and had to try it with a rather poor idea and limited time. I regret not being able to prepare better, but if there’s another chance in the future, I’d like to come back with a proper idea.