An attempt of a refactor into something vaguely similar to a social media post...

At unyt.org we believe in public goods and the importance of community-driven development, therefore also in the concepts of open source and transparency.

Our entire idea of open source is based on the concept of people collaborating - sharing their work - creating software that actually benefits, not just profits. But with the advancements around LLMs (a.k.a. "AI"), the concept of community-driven development seemingly drifts away from benefiting everybody to profit a few.

Sadly we can't avoid noticing how easy it is for an LLM to generate - not to say "create" - some software by just replicating existing source code on which it was trained - or which it can easily find online... commonly without giving any credit to the original authors or contributors.

A recent example which demonstrates this problem is the case of Chardet, a popular Python library for character encoding detection. It was completely rewritten using LLM agents and then released by the current maintainers - under a new license without giving any credit to the original authors or contributors.

This sets a very dangerous precedent of how LLMs can be used to undercut the very principles of open source and community-driven development.

Besides from obvious rewrites, there are also demo projects like the C compiler that was recently published by Anthropic - claiming that it was fully generated by an LLM.

While this definitely is an impressive achievement, it is questionable whether it is reasonable to push existing, high-quality, human-created code into an LLM to just reproduce the existing. We don't like to blame tools - but maybe their use-cases.

Completely disregarding ethical or legal questions when it comes to copyright and license of intellectual property, this situation becomes a major problem, because it is way too easy to create lots and lots of low-quality code, which nobody ever felt, feels or will feel responsible for.

Since we have lots of contributors and interns who are just starting out with programming, we want to ensure they get the opportunity to learn and grow as developers. We don't want to invest time into our junior developers to just be able copy-pasting blocks of code from an LLM without any understanding.

Especially for our interns we consider it critical to learn how to write code by themselves so they get a fair chance to understand the fundamentals and principles of software development.

Furthermore for more experienced developers we consider it important to maintain the quality of our codebases - by ensuring contributions are human-related and someone takes actual responsibility for them.

This is why we at unyt.org came to the decision to keep our software human-created by enforcing an AI-policy for all of our projects.

We are not anti-AI in general - we actually encourage the useage of AI tools which assist developers to solve problems or jumpstart research and alike - but we strongly oppose the usage of fully automated, agent-generated code in all of our projects.

Aware of it being impossible to fully determine whether a piece of code was generated by an LLM, we will reserve the right to disregard any contributions where we have reason to believe were generated by an LLM, as well as the right to remove any code from our repositories that we have a justified reason to believe was generated by an LLM.

LLMs are definitely fascinating and can be a very useful tool in many use-cases, but they are not a replacement for the creativity, innovation, or collaboration within the open source community.

Thank you for your attention,
in case you want to support our cause we'd like to invite you to visit our merch store at ...

In case you'd like to become our next intern - in around four weeks we'll be looking one - thus there is enough time to polish your cv and update your public github. Good luck!

Enjoy your day. 🌺