Fucking tmux is AI slop. #openbsd #tmux #floss https://github.com/tmux/tmux/wiki/Contributing#use-of-ai

EDIT: There has been a fair amount of discussion about if this applies to upstream openBSD, and how the AI code policy is interpreted.

An example can be seen in this commit by the lead developer landing in openBSD attributed to "claude code" https://github.com/openbsd/src/commit/9c2b8e445a0bdfafdd6148b1760f00aa5429627b

Contributing

tmux source code. Contribute to tmux/tmux development by creating an account on GitHub.

GitHub
losing one tool after another

(T_T)

i really hope that slop doesn't make its way into #openbsd's base. lately it feels like i'd have to fork everything i care about...

@hi @trashheap
Seems to me a bit more nuanced:

Use of AI

Code written with the assistance of AI can be acceptable. However, the question of ownership and copyright of AI-produced work is not yet well-defined in law.

Given this, in order for code produced with AI to be accepted, it must either be trivial enough to be not copyrightable (basic refactoring, one line bug fixes), or there must be a public statement available from the AI publisher showing they do not assert copyright over the work.

@ParadeGrotesque @hi @trashheap
That last part shows a profound misunderstanding of the problem.

@FritzAdalis

Why do you say that?

Seems to me a cautious statement, in line with the rest of the position.

You could even understand it as a way to say: release the code or state clearly that it is free of problematic license.

Which is another way of saying: we won't accept your code unless you can vouch for it. Which no slop shop is able to do.

@hi @trashheap

@ParadeGrotesque @hi @trashheap
I don't think anyone is worried that Anthropic et al are going to assert copyright; it's more that copyrighted works were used to train the model so the generated code may or may not fall under the original authors' copyright.

How can you vouch for code you didn't write?

The various other AI concerns are externalities I guess.

@FritzAdalis

And I believe that's exactly the point the tmux author is trying to make: you did not write the code, you cannot vouch for the code, therefore I can't accept your slop generated code.

(Unless it is completely trivial, in which case your code is probably not needed)

@hi @trashheap

@ParadeGrotesque @FritzAdalis @hi

You could even understand it as a way to say: release the code or state clearly that it is free of problematic license.

Take a look at that bit of text again, specifically the text which FOLLOWS it.

It is followed by a listing of LLMs who have issued statements from various LLM vendors that they assert no copyright on the output; and are therefore "fine."

@ParadeGrotesque @FritzAdalis @hi

And I believe that's exactly the point the tmux author is trying to make: you did not write the code, you cannot vouch for the code, therefore I can't accept your slop generated code.

As an example you can see a fairly lengthy pull request "co-authored with Claude" here: https://github.com/tmux/tmux/pull/4744/changes/b700e9ce219cae63988c4287fd3cde41a6a6f8c4

AND then you can see it landing in upstream tmux in the openbsd source tree, attributed to the original author and claude here: https://github.com/openbsd/src/commit/9c2b8e445a0bdfafdd6148b1760f00aa5429627b

The ultimate committer of that code being Nicholas Marriott, the lead developer of tmux.

Support synchronized output mode (DECSET 2026) from applications by chrislloyd ยท Pull Request #4744 ยท tmux/tmux

When an application enables synchronized output (ESC[?2026h), tmux now defers flushing pane output until the application disables it (ESC[?2026l) or a 1 second timeout expires. This reduces tearing...

GitHub

@trashheap

OK, that's bad. As in: really bad. โ˜น๏ธ

@FritzAdalis @hi

@ParadeGrotesque @FritzAdalis @hi YEAH, it stings; tmux has been part of my personal stack for a long time.