this post was submitted on 16 Aug 2026
570 points (98.3% liked)
me_irl
8187 readers
1127 users here now
All posts need to have the same title: me_irl it is allowed to use an emoji instead of the underscore _
founded 3 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
The only way to use it for code is to spend exponentially more time than it would have taken you to just write the code. You basically have to just take the results and bounce them back and forth between different models and be like correct this code back and forth in an endless loop of different models until it finally gets something that's actually pretty good.
But the amount of time you spend doing that and getting it to actually put out something useful you could have just done it yourself and then went and played some games with the extra time
You're assuming the person using the AI already knows how to code.
I don't. I basically don't even know how to use Linux beyond installing it. And yet I now have Codex running on an always-on laptop, with access to my Plex server, that I use whenever something goes wrong.
When there's a problem, I just tell Codex "this isn't working" or "I can't find this thing I expected to be there." It goes and inspects the system, figures out what's wrong, changes whatever needs changing, tests it, and fixes the issue. Then later I check Plex and the thing I was looking for is just there.
I'm obviously still the one telling it what I want and checking the result, but I'm not bouncing code between five different models until something happens to work. Most of the time I don't even know what code or commands would need to be written in the first place.
For me, that's actually the biggest thing LLMs are doing: they're reducing the friction I've always had between what I wanted to do with technology and what I was actually capable of doing myself. A lot of projects or fixes that previously would have required me to learn a bunch of coding first are now things I can at least attempt immediately.
So "you could have just done it yourself faster" doesn't really apply to me. I couldn't have done it myself faster. In many cases I probably couldn't have done it myself at all without first spending a ridiculous amount of time learning the underlying stuff.
And I'm not here trying to advocate for ChatGPT or convince anyone that AI coding is amazing. I just had to reply because what you're describing is so completely different from my experience using Codex this past month.
Reply drafted by Chatgpt by the way, hope that doesn't make anyone mad, I still took the time to proofread it and make it mine.
What you have described is not coding. You did not create an application, you had it manage a media server. That's systems administration and yes llms are generally pretty decent at that because the documentation is what they were trained on and it's all very straightforward if this then that type things.
When we say AI coding we mean actually creating an application from scratch. It's a very abstract concept with a crap ton of rules and llms are still very bad at it
You're right, but I also use codex at work to create Chrome extensions from scratch that automate processes.
I see the limitations and have encoutered them trying to develop more complex programs for work.
It's great for creating chrome extensions or scripts that help me tremendously at work. It's still not good enough to let me, a complete beginner, create an app for work that would replace a service my boss pays 100€ a month for. (For now at least, I haven't given up).
I'm also trying to develop a Deltarune like game on Godot using Codex, to see how far the tools currently go, I don't intend to publish o montetize it of course it's just a passion project among friends.
One year ago trying to use AI to setup Plex completely wrecked my mini-PC and I had to reset it, now it does everything by itself and I just supervise.
I wouldn't bet against it is all I'm saying.
I mean even as it improves I'll be against it for the same reason I'm against higher level languages. It creates so-called developers that have no idea what they are doing no idea how the code works and no idea what the actual mechanisms behind it are. We've done that with web developers treating them like they are actual developers when they have no idea what's going on and it has led to a situation where just a simple goddamn weather widget needs more than one gigabyte of ram to function. Where as an actual genuine developer that truly understood code could make the exact same thing and use less than 25 MB AI assisted coding is only going to continue to exacerbate this issue with developer knowledge the exact same way upscaling in video games as exacerbated the problem with developers that don't know how to optimize their game. We are already seeing games coming out that are dramatically worse at optimization and they just expect you to use upscaling as a Band-Aid to fix it and make it vaguely playable
A calculator lets people get useful results without understanding how arithmetic is implemented internally. That doesn’t make calculators bad; it means the required expertise depends on the stakes. I wouldn’t want someone designing a bridge who doesn’t understand engineering. But I also wouldn’t tell someone calculating a grocery discount to learn numerical methods first. That’s how I see AI coding: if I can make a small tool that solves my problem, test it, and verify it works, “you don’t understand the implementation deeply enough” isn’t automatically an argument against the tool. Otherwise you can make the same argument against basically every abstraction in computing.
I see your argument and I think there’s value in the concern. AI absolutely can let people build things they don’t understand, and that can become a problem when the stakes are high.
Where I disagree is treating that as a reason to oppose the abstraction itself. I think you’re focusing entirely on what gets lost when the barrier to entry drops, while ignoring what becomes possible for people who previously couldn’t participate at all.