Why This Matters
A developer's rant about receiving a 5,000-line, AI-generated pull request for a payments backend that his partner never even ran illustrates a growing friction point in software teams: AI can produce code volume far faster than humans can review or verify it. The story is less about AI capability than about accountability — skipping basic testing and reading your own docs shifts the cost onto collaborators. As vibe-coding spreads, code review, trust, and ownership norms are becoming the real bottleneck.
Key Takeaways
- A 5,236-line AI-generated payments PR was submitted untested and simply didn't work when the endpoints were run.
- AI-written documentation padded obvious details (like 'returns 400 on error') while omitting essentials such as where to get API keys.
- The author frames heavy AI reliance as a skills-atrophy risk, arguing verification and human collaboration can't be delegated to the model.
My business partner sent a 5K vibe-coded PR that he didn't even test
08 Sep, 2026
The whole payments backend module, which I specifically told him to work together with me (I do frontend), was vibe-coded by him in one single day.
How nice, to not have to discuss with the human being who will be using your code and knows what he wants, and instead work with a sycophantic "yes man" AI right?
All 5,236 additions to the Github PR, including a documentation that is so blatant of AI redundancies (for example, saying that an error returns "400" on the API).
I understand that AI doesn't know that I know what a 400 is, but my friend should at least read that document and know better.
So here's the sad part. I understand y'all vibe-coding features and being 10x engineer and way faster than me.
But I just went to test the PR...
It doesn't work.
He didn't even TRY the endpoints.
... continue reading
GoKawiil