Because they make loads of code that looks correct (it's LLMs) but has issues.
this post was submitted on 20 Jan 2026
52 points (96.4% liked)
Programming
24562 readers
112 users here now
Welcome to the main community in programming.dev! Feel free to post anything relating to programming here!
Cross posting is strongly encouraged in the instance. If you feel your post or another person's post makes sense in another community cross post into it.
Hope you enjoy the instance!
Rules
Rules
- Follow the programming.dev instance rules
- Keep content related to programming in some way
- If you're posting long videos try to add in some form of tldr for those who don't want to watch videos
Wormhole
Follow the wormhole through a path of communities !webdev@programming.dev
founded 2 years ago
MODERATORS
I'm not reviewint that butload of code. Just today I had to rollback some shit I "reviewed" a month ago. Huge ass refactor. I reviewed for 1 hour and then I just yolo'd it.
I'm not reviewing anymore.
They want to decimate software engineering because after unionized government employees have been obliterated it’s the last well-paying job where employees sometimes feel bold enough to take principled stands against management.
It started with Visual Studio — the concept was to make programming “easy” enough for a manager to do it and cut out software engineers. Then it was off-shoring. Then it was coding bootcamps (note it isn’t software engineering bootcamps).
Now it’s AI.
The claims of efficiency or productivity are baloney. Companies demand RTO because it’s more satisfying for middle managers to throw their weight around IRL, not because commuting two hours to sit on Zoom calls is more efficient.
They care about control, not efficiency or productivity.
Thankfully we’re really good at reviews. Right?
Oh everyone absolutely loves reviews to bits
I'm fine reviewing, if I'm familiar with the task, codebase and especially if I've been involved with the proposed technical solution.
reviewing gen code? Hell, NO!
I believe for simple, common tasks and small projects, it's at least manageable. Even good, if you are not forced to use AI and can simply choose to what extent you use the tools and its results.
But with enterprise-level coding agents that are supposed to handle issues in long-term projects, the work is shifted to every other phase really, before and after the coding. Initially I thought that proper documentation at least benefits everybody, but prompts and instructions optimized for AI are not necessarily good documentation for humans to read.
Yep.
Why do you want code?
: To do task.
Why do you want less code?
: Because ca. every 1000 loc has a bug and more code creates more maintenance cost.
Why do you not want the most perfect code?
: Because it's more dev work, cost's more money.
So what do you want in a software developer?
: He should balance the required work (in loc and in time) to do task.
What does LLM do?
: Creates a lot of loc to do task.
What does LLM not do?
: No balance. No quality control. Only loc.
So what do you get with LLM?
: More loc. More bugs.
Do you not want that?
: No, it creates a lot of maintenance cost.
: Plus, fixing someone else's code is harder than writing your own, especially code that on first look seems ok; more burnout in devs.
Edit: seems like lemmy has no definition list support?
Big agree.
But also: people seem to only focus on the output side of the task of writing code, and forget that the developer also receives input from the codebase in return.
Even if you end up with exactly the same code artifact after completing a work item, you’ll have a better understanding of the codebase without delegating swaths of it to AI. But bosses tend not to consider this.
Tech bros have successfully convinced people that mental states do not exist, or at least do not matter — for laborers, anyway, cuz they’ll happily claim that their superior thoughts are exactly why they deserve to be billionaires.
Let the AI do the review as well.