If you're a programmer and you feel depressed by AI, don't be!

Published on 2025-10-08. Modified on 2026-05-10.

If you're a programmer and you use AI, my advice is to use it sparingly and treat it as nothing more than extended search, a tool that makes many mistakes.

Let me first clarify that even though I believe that the term Artificial Intelligence (AI) is perhaps the least justified term to use, because in no way whatsoever is intelligence a part of the technology, I have decided to use the term in this blog post for simplicity. By the term AI I refer to any kind of related hyped up technology, whether it's LLMs, artificial neural networks, adaptive algorithms, so-called machine intelligence, etc.

Then let me begin by citing someone who shall remain nameless.

AI is writing 90% of the code now!

― Some bloke

To which I will respond with the following comment:

I am sorry, I don't want to offend anyone, but I don't care for how long you have been programming, if AI is getting you 90% of the way, either you live on a completely different planet than I do, or you simply cannot be a very good programmer. Already too much software and too many websites suck beyond comprehension and you're not making it any better!

With that said, let us get something established right away. I believe that there is only two types of "programmers" who are exited about the code AI produces. A person new to programming, because he or she doesn't know any better, and then the person who doesn't care much about coding or programming at all, someone who (more or less) only cares about shipping products. That person, the one who only cares about shipping products, are also the absolute worst kind of IT manager. They are a "cancer" to the IT industry. I call them a cancer because of the direct damage they do to the software industry and the Internet in general.

There are such types of work which simply cannot be just about shipping products and earning money and when it is, the result is always only bad. When software is bad there is almost always a negative consequence, relatively speaking of course, but nonetheless.

The AI hype is in my opinion the thing that has caused the most damage to the industry ever. And the reason is simple. AI cannot - and it will never be able to - deliver the results that have been promised.

Google's CEO Sundar_Pichai has stated that AI could be more profound than electricity or fire. When he said that, he was not just selling a product, he was not just exaggerating, he was flat out lying! And the same goes for all the other Big Tech CEO psychopaths.

For the uninitiated, AIs are basically artificial neural networks which are governed by mathematics, algebra to be precise. The process called training, which involves exposing the neural network to data again and again while gradually adjusting the weights of the network nodes, is not only filled with pitfalls but it also reaches a certain point at which you simply cannot get any better results. And we have already reached the peak of the technology. The amounts of improvements that can be made will only produce very small differences from what we are already seeing. This is a basic limitation of the fundamentals of the technology.

However, all the enormous amounts of money that Big Tech have managed to get invested into this massively over-hyped technology needs to be returned to the investors with profit, which is not going to happen in most cases, but Big Tech and the media will keep the hype running as long as they can because the hype is the product.

I am a huge believer in AI, I am not a huge believer in the whole, things going on around the AI. I find the marketing and the market to be sick, twisted, and there is going to be a crash, and it's going to be ugly.

― Linus Torvalds

Some say that AIs are empowering regular users to build and customize software that improve their lives. Everyone seem to overlook the fact that AIs are too expensive to build and run by individuals and even smaller companies hence people depend on either payed versions or public free versions from Big Tech, such as e.g. Google, Facebook or Twitter/X. Who knows what these AI versions contain? How they use the data they obtain from usage? Already Big Tech have abused and violated peoples privacy and other rights. These AI tools cultivate the usual centralization and dependency. We need to move away from that, not embrace it!

The programmers I know who care about producing good software share a common perception about AI after having tried out the different popular models, they can get you 15% of the way, perhaps 20%, but mostly they are just not worth the effort and they generally produce bad code.

The first examples I have seen has not really been promising. That is, you get something done in maybe a 10th of the time, but it also has three or four times as many bugs. And now you have to find those bugs and remove them. But that is tedious work. And that is exactly the kind of work you're trying to avoid.

― Bjarne Stroustrup

Some studies e.g. also show that when developers use AI tools, they take 19% longer than without. This is mainly related to experienced developers and I personally believe there is a good reason for that, you spend more time scrutinizing the code and fixing basic mistakes.

The other problem is that you need to train those models. And where do you get the training material?
People are using GitHub for instance, getting all the stuff from the open source community. The problem is that a lot of that is low quality. Sort of student level kind of code, and the really, really high quality professional code is not always there. And even if it is, it is drowned out by the endless student projects.

If you go on to GitHub, what fraction of the stuff is actually maintained and used, and what are used on scale other than (just) ten people?

So where do you get the training material?
I think, if you could get really good training material, for the area of programming that you are interested in, you could get a lot of help from the AI. But now we're back to actually where we started, which is programming is not one thing, you don't write the same kind of code for my brakes as you do for my website for foodstuff.

And if you just go and get all of this data in for the AI, you're likely to sweep across all application areas. You're going to sweep across (different) sort of levels and expertise.
Maybe they will figure out how to solve that problem. I certainly hope so. But as of now, I see that as problems and I am not at this stage going to encourage use of those techniques in the domain that I am interested in.

― Bjarne Stroustrup

I don't actually see it building a simple piece of code that is good enough up to the best standards today.

― Bjarne Stroustrup

Some companies have also finally realized the hype and have begun reverting back to how things where before they stated using AI. And about 95% of companies that try AI aren't making any money from it.

The fact of the matter is that the benefits the technology can provide are limited to very specific tasks and software engineering and programming is just not one of them.

Image recognition, sure, you can get amazing results. A drone can fly by itself, follow you around from up high, steer clear of physical obstacles, etc. And that is exactly because of the artificial neural network and the mathematics behind the calculations of the weights. This is extremely useful for pattern recognition in the form of images.

To some extent, spoken language as well, as long as we're dealing with basic grammar and translation. But not much beyond that. And definitely not when dealing directly with the work of a linguist.

Now, if you love programming and you feel depressed by AI, don't believe all you read and hear. Keep programming with a passion. Do the projects you care about, the projects that gets you exited, and simply ignore what every bandwagoner is saying or doing.

Some parts of the industry is changing, yes, but the part that is undergoing the change is not changing for the better and it is that specific part that you should stay clear of anyway. It's the bad part. So if a company requires you to use AI for software development, use that as an indicator that this is probably not the best place to work.

Perhaps important lessons will eventually be learned and the industry will change for the better after they have been bitten severely in the ass when this one crashes. One can always hope, right?

In any case, if you're a programmer and you use AI, my advice is to use it sparingly and treat it as nothing more than extended search, a simple tool that makes many mistakes. Always combine it with regular search and do the research and experimentation you would normally do. You need the process of normal research, critical thinking, finding different solutions and suggestions, looking through different code examples from other people, who, by sharing their code also share their experience, reading peoples comments and getting feedback (from people not AI), trying out different stuff and approaches, otherwise you will wither and you will not grow. And perhaps your efforts will also effect how you define yourself.

推荐订阅源

Hacker News - Newest: "AI"

Relevant reading