AI turned me from a 1x developer into a 0.1x peer reviewer

Link copied

Published

I am now the blocker. The junior just pushed 4 PRs of code that looks PERFECT and require 110% of my brain power to review.

There are new org-wide meetings for learning about how other areas are using AI. These are almost always demos for greenfield solutions to problems nobody has, OR they are desperate attempts to paint over 20 years of tech debt with a new Next.js dashboard with floating, fuzzy color gradients that barely run on a top-of-the-line MacBook Pro.

I am ostensibly the "architect" and "lead developer" for many disparate applications and 90% of my job is now sifting through perfect-looking, thoroughly FUCKED code that I know the junior is smart enough to write SOME DAY and that he absolutely does not write now. The more scared devs will at least say "This is from the LLM but it seems to work" in their PR, but most of the time you don't even get that disclosure.

The contractors are more mercenary than ever. I just got tagged for review in a PR with a 3,000+ line diff that touches the database schema, back-end, and front-end, and fails in the test pipeline. "I need this merged in time for the demo." Cool man, I guess I'll go fuck myself.

Let's deploy the new tech debt

It used to be, you could tell a repo was bullshit when you dropped in and the code formatting was off, comments were non-existent, the variable names bore no relation to the what they referred to, the README was barren, and everything pointed to nobody giving a fuck about the thing.

It turns out, those signals are all totally valid and valuable. If a thing didn't matter, it looked like it didn't matter. You could judge how important a thing was based on a few seconds glancing around.

We're in a whiplash world now, where at work, I see repo after repo popping up with tens of thousands of lines of code, well-manicured READMEs full of lists bulleted by emojis, and the most fucked, non-running code I've seen in my career. Genuinely, I don't believe people are even running the code before committing and pushing it now. If you try to simply install the dependencies listed in the manifest, it won't complete. Never mind trying to run or compile the thing.

Obviously all the new stuff reeks to high heaven, but it's presented with the same (or greater) level of importance as the most vital repositories in the whole organization that have been carefully crafted over years.

The only signals left are horrible biases - "I know this particular retard is a slop enjoyer, so I can't trust anything he submits anymore."

It could be worse, I guess?

My org is old and crusty. Most people are extremely concerned with the ethics, environmental impact, and general rot that AI is causing. It's truly not even that bad here. Speaking to a friend in the IoT space was a real eye-opener.

At their small org, the CEO and CTO are both extremely adept architects and coders and completely bought into the slop stack, top to bottom. Times are tough everywhere, and they recently laid off several high performing developers. Meanwhile, everyone else in the org has been told to adopt AI or die. Everyone has their token usage tracked. No purple Nikes, yet.

In their most recent demo day, the HR lead demo'd an app she built and launched that replaces their multi-thousand dollar annual SaaS bill for software that was used to generate an org chart.

(If you're wondering why the fuck a small IoT company is bothering to spend ANYTHING on software for org charts, you are seen.)

The CEO was thrilled. He even helped her get the app from localhost to cloud for the demo. Progress!

The CTO nearly shit himself, realizing the entire employee database was now exposed to the web, without so much as an HTTP Basic password in front. Just an open API and 10 trillion Tailwind CSS classes. Vibes, man.

Naturally, whatever savings might have occurred have been eaten up by the CTO having to go add security to the thing. From an optics point of view, it was really important to praise the AI usage and not bring up the security ugliness to the group.

I'm sure whichever dev gets the ticket to go fix that thing when it breaks will feel good that they are in such a forward-looking org.

Times are tough and a bad job is better than no job

I guess if my job is now Head Pull Request Reviewer and I no longer get any say in architecture or development before the code is written, that's fine. Just keep the checks coming. But I'm turning off PagerDuty since it no longer matters if our stuff is up or not. The customers can eat shit. The AI will fix it.

It's 17:00 somewhere.