---
title: "Is built with AI an insult?"
description: "The vibe coded label is earned when nobody reviewed the code, but the line is whether someone understood it and answers for it."
url: "https://michael.zazon.us/notes/is-built-with-ai-an-insult"
published: "2026-11-01T12:00:00-05:00"
kind: "take"
subject: "ai-agents"
tags: ["ai-coding","code-quality"]
---
humans shipped bugs for 50 years. the difference now is we get to catch them faster

But "vibe coded" has come to mean [a label for anything AI touched](https://en.wikipedia.org/wiki/Vibe_coding), instead of what it originally meant: not looking at the code at all. [Simon Willison draws that line clearly](https://simonwillison.net/2025/Mar/19/vibe-coding/), saying that if you reviewed and understood the code written by the LLM, it doesn't count.

AI is a tool; what you produce with it is your responsibility, whether it is good or bad.

Every company I have worked with over the last few years from startups to the largest banks in the world are building and shipping products that delight their users at record speeds using AI assisted software development and agentic coding.

In my career I have written code, reviewed code, and reviewed and approved and rejected PRs all by hand in the past.

Reviewing and rejecting has always been my job. The agents didn't take that away. It's just that now the *author* of the change isn't me, but the *owner* is still the same person it always was: me. (see [Who owns the code Beyond Vibe Coding held up against 2026](https://michael.zazon.us/notes/who-owns-the-code-beyond-vibe-coding-held-up-against-2026))

And yet most people are doing that.. prompting and praying. That is exactly why there's such a negative stereotype on ai coding, and the ai hate isn't irrational because most people don't know how to use ai correctly.

I only have issue with AI-generated garbage posts that clearly have not been even read or edited by the OP. If they can't take the time to read and edit what AI spits out, why should we?

It's difficult to argue against this. A lot of AI-generated content is being published without the author even reading it, and it shows. Unfortunately, this label tends to stick because it's often well-deserved.

This keeps the agent code *mine*, because I can put up review gates and keep the interfaces small. Automated reviewers can stare at diffs forever, but they can't know what the purpose of a change is (see [Every PR has a crowd of reviewers now and none of them know why](https://michael.zazon.us/notes/every-pr-has-a-crowd-of-reviewers-now-and-none-of-them-know-why)). That's still my job.

The issue is not the AI usage, its what follows. Lack of understanding, lack of knowledge, lack of competence in the end due not using those skills anymore..

This is where I'd draw the line. I'd judge whether or not someone understood the code and answers for it. It's the same question as "who gets paged when it breaks?" which is at the core of [The AI argument at work is really about who's on call](https://michael.zazon.us/notes/the-ai-argument-at-work-is-really-about-whos-on-call).