AI Makes You Think Your Code Is Better Than It Is

Описание: I reviewed AI-generated code in my Vue app and realized something uncomfortable:
the app worked — and that’s exactly why I trusted it too much.

In this video, I walk through parts of my codebase that were generated with the help of LLMs - what people often call “AI slop”. Not broken code, not useless code, but code that looks fine, works, and quietly lowers quality if you don’t review it properly.

The problem wasn’t that AI “broke” my app.
The problem was the false confidence it gave me.

This isn’t a rant about AI.
It’s a responsibility check.

If you’re using AI to speed up development, this is what reviewing AI-generated code actually looks like - including duplicated logic, questionable structure, and those “this feels wrong” moments that only show up when you slow down and read.

I’m a self-taught developer, not an enterprise architect. This video shows my real process, not best practices on paper.

---

Timeline

00:00 – Intro: what went wrong and why this matters
01:40 – AI slop in types
03:43 – Duplicated class logic
04:59 – Background info and missing interfaces
06:23 – Using Opencode for boilerplate
06:41 – My mistake in types
07:46 – Boilerplate is a good use case for LLMs
08:51 – A good LLM suggestion: paginated types
09:30 – Why, AI, why? Imports and type placement
11:30 – Review your code: AI likes to define types everywhere
13:36 – Rant: redefining the same types and AI lying about imports
15:40 – Types done: what’s next and what isn’t
16:48 – UI issues: inconsistent Tailwind design
18:04 – Unused HTML artifacts
18:57 – More considerations and inconsistent AI design
21:35 – Summary of my findings
22:18 – Next time and outro

---

Topics covered

Reviewing AI-generated Vue / TypeScript code
What “AI slop” actually means in practice
Where AI output looks fine but isn’t
Why “working” ≠ “good code”
Skill issues vs tool issues
How I approach fixing AI-generated slop