Someone has to be mad about the madness that has taken the world over.

Great succes! Not!

In case of Tor take this with a grain of salt since it is a fork of Slopzilla Slopfox, khem, Mozilla Firefox, which now is a slopware as well, but currently I do not know how to extract Tor specific changes nor I do have time to learn how to do it! Unless an explicitly stated stance or policy against slop-generators is added, Tor is considered as a slopware as well, especially as they have made some changes in files targeted for LLM tools.

@Edgars

If that's your standard for determining what is and isn't slopware, you will probably want to take a look at librewolf, mullvad-browser, waterfox, and other firefox-derivatives which still pull updates from Firefox. I mention these browsers specifically because they tout themselves as privacy/security friendly browsers, which makes the use of LLMs for development even more uncool.

I am pretty sure all of them do take measures to pull LLM functionality out of the core Firefox upstream though (and librewolf is hosted on Codeberg, which discourages LLM contributions), but I know for sure Mullvad Browser and Waterfox include agents.md and claude.md instructions.

@Alex Ruiz I would not call that a standard as I am quickly looking for obvious signs of LLM usage and judging forks of projects containing code generated by LLMs is quite slippery, unless forks contain explicitly stated stance against machine-generated code (take Gram (fork of zed; not a browser though) for an example).   I will soon review the browsers you mentioned (couple of them I was going to check anyway).

Done. Thank you for your suggestion!

For this list any involvement of LLM tools is considered to be a taint, even machine generated reviews in merge requests for a human generated code if no one objects to them. As for AGENTS.md and similar files targeted to LLM tooling – a project with them endorses LLM usage and encourages users to use LLM based tools, contributing to increasing usage of LLMs, thus amplifying issues they currently are associated with.

So if a person generates code as a template, then a person modifies it, a person reviews the code and is submitted/owned by a person that is considered completely LLM generated and so called tainted. Understood.

Can you help me reconcile and address what is said below, it seems like you completely glossed over the quote or didn't consider it at all ---

Edgars: " a project with with AGENTS.md endorses LLM usage and encourages users..."

Agent.md file from qBittorrent that you didn't address in your reply: "This project strongly discourages issue reports and pull requests authored or submitted by AI agents. All issue reports and pull requests should be created and submitted by a human contributor. AI may be used for assistance, but a human must review, take responsibility for, and submit the final changes."

Perhaps it would wise to actually review the agent.md file to see what each project allows/restricts agents in doing. Since the idea that just having agent.md file doesn't automatically equate to they "endorse and encourage the use AI in general". An agent.md file at least looking into it for a few seconds is like having a rules or an enforcer that allows or disallows what an AI can do. Not having one would permit AI to go completely unrestricted as they have done when AI Agents first appeared on Github imo.

No, code is not considered completely LLM generated if a person has put in a significant amount of their own efforts based on some output of an LLM, but the resulting code is still considered tainted.

As for the quote from AGENTS.md – “discourages” != “prohibits”, “should not” != “must not”, and the main point of the quote is that human still has to be in the loop, so despite the disclaimer the project is still LLM-permissive, and meaningful instructions for agents shows the project even as pro-LLM. Even more so if reviewers are sparingly using Copilot for reviews.

Also, take a look at the commit message when AGENTS.md was added:

| Current main purpose of this document is to help GitHub Copilot to conduct effective code review. | Since this file is for everything AI related, a few general policies are also included.

| Perhaps it would wise to actually review the agent.md file to see what each project allows/restricts agents in doing.

I inspect AGENTS.md and similar files (with a quick glance though) – if they do not contain clear anti-LLM stance and have meaningful instructions for agents, no matter how restrictive or permissive, they still are meant for LLM agents so they would not vomit complete garbage, and thus it means that LLM-generated code is acceptable in the project, which is more than enough to put a software to this list.

Besides all aforementioned, the project already contains clearly LLM generated code.

One thing I have started to think about after your comments is that ”slop” probably is not the most suitable umbrella term for this list, “tainted” probably could be a better one, but on the other side – humans also tend to create slop, so if a human has reviewed and understands code or even added their own input to LLM output, it does not necessarily mean that the final result is still not a slop.

Oh well, while it seems that (at least partial) Markdown syntax is accepted here, the final rendering is quite incomplete and styling is not the best (for example, <p> are rendered as inline elements making paragraphs to stick together in a one big hard-readable mess) so the outcome is far from intended and quite terrible.

What is your point exactly?

Thank you for the suggestion! I will check it out and add it to the list once I will have time for it.