
For more information on why we think this list matters, please see out Why not LLMs? file.
Want to add, update, or remove a project listed below? Check out our Contributing Doc.
See a project on here that you love, and wish wasn’t on here? Check out our Guide for Removal and/or No AI Request Template.
Ready to fork a project to introduce a No AI policy? Check out our How to Slop Free Fork Guide.
Any other questions? Please check out our FAQ, and if your question isn’t answered there, please feel free to start a Discussion.


So one of your main points is that writing code with LLMs cause people to understand their code less.
And instead of trying to fix this (as we engineers always do in such situation) and build tools that will actually make understanding easier without sacrificing benefits of LLMs give us, you are just crying to stop using entire technology.
There is no coming back. I liked writing code too, but it does not make sense to do it all by hand anymore.
I don’t think you’re truly understanding what you’re suggesting here. If such tools could be made, they would indeed reduce the harms caused by using LLMs. They would also revolutionise education.
What you’re essentially expecting from people is that, instead of working on a list like this, they challenge tens of thousands of years of history of people thinking about how to best teach other people, and come out victorious. All to slightly reduce the problems caused by having an LLM generate code while still offering no benefit.
Anyway, if you have any professional ethics at all, writing code by hand is the only thing that makes sense.
I’m not sure we’re on the same page. I’m not talking about revolutionary superhuman way of putting understanding into humans heads.
I’m a programmer by profession and I do confirm that mental fatigue when using LLMs is a real problem. But I believe it is because how software that we use to interact with LLMs works. It’s because LLMs tend to do many thing at once which in turn causes programmer to also focus on many things simultaneously. It’s because agentic software promotes multitasking and it promotes it because generating an answer often takes time.
So what I’m suggesting is that we should try to develop a more “human-friendly” way of interacting with LLMs. A way that breaks the decision process into smaller and sequential tasks, but still making programming syntax less relevant than when we coded “traditional way”.
That just gets you closer to just writing the code with worse syntax, though. The reason for your mental fatigue is that you’re spending all day trying to convey concepts that are not intuitive to describe in natural language… using natural language. It’s like you’re in a meeting with your boss trying to explain the cause of a bug and why it’s so difficult to fix, all day, every day. And the smaller the chunks you have it generate are, the less intuitive it gets to explain what it needs to do; higher level concepts are usually easier to explain than lower level concepts. So it would, if anything, be less human-friendly and still not come close to writing the code yourself when it comes to carving intuition for it into your neurons.
I mean, no one is forcing you to be part of this. It’s not like when we created photography we stopped painting portraits. I like coding and I’ll keep doing it. It’s less efficient? Yeah, but I understand my code better and I constantly learn new things. Whenever there is a problem I know exaclty where to put my hands to fix it and I have fun doing it. Paying tokens so that someone else takes away my fun sounds kinda stupid. I’m not asking an llm to solve my sudoku or crossword puzzle…
Edit: sorry, missed an important point. Why would someone want a community of hand made coding you may ask. Well, if you want to interact with people that actually code and can answer your questions on they code they write you kinda need one. Otherwise you are just talking with an llm through a human interface
This is the same core argument as the people who clamor for revolution. However when pressed, they always go, “Well, we don’t have enough momentum, you see. It’s not that simple. We need more time to get enough people. Etc.”
People like OP are actually doing something. Instead of punching down, show us what you’ve been doing to combat AI psychosis or LLM-induced cognitive decline. Show us the viable alternatives that have been created with human flourishing in mind. Which engineers are working this problem?
We are talking as if cognitive decline is some proven fact. If I’m designing a webpage and instead of spending time figuring out “how to center a div” properly or similar nonsense, I’m actually spending time at thinking how my webpage could be actually useful, do I have cognitive decline? If I don’t memorize Terraform syntax anymore, because I can have it written the way I want automatically, and I spent more of my time designing app’s architecture, do I have cognitive decline?
Science doesn’t prove things. However, the studies do in fact indicate that LLM-induced cognitive decline is a real thing, and while you are correct that it’s not considered a fully settled question (in no small part because AI-use is not purely good or purely bad), the way models currently function consistently indicate a decline in cognition, particularly with regular use.
Maybe, but I honestly don’t know, and frankly neither do you. However, offloading your cognition to an LLM does seem to be a pathway to cognitive decline.
That’s like adrian zenz talking to 8 people and then saying “this satellite photo of a school is proof of genocide”
lol your comparison with the revolution is perfect. of cpurse it was a liberal who supports reactionary movements.