PR submission slowdown for the coming few weeks

[ann0see]

We would like to focus on reviewing the large amount of new PRs due to the influx of AI generated changes. This needs some time. Therefore, please open issues for defects and not PRs. These issues may include diffs. In case there is something important feel free to open PRs - however we appreciate if you waited a while until we've cleaned up the inbox a bit.

In the meantime:

• Please self review your changes on your own fork first
• Closely assess if the PR follows CONTRIBUTING.md. Ensure that it is reviewable
• let an AI review your changes and fix them if you wish so. Opencode has free models which can review your PRs in a decent quality. This exists for every major desktop OS

[Tomosu-AI]

With the amount of AI-generated PRs coming in, another thing contributors could try before opening one is a quick PR check, especially for dependency, config, CI/CD, or production-facing changes.

We’re building Tomosu around exactly that: giving you a risk signal before a change ships.

Might be useful here if anyone wants to give it a try.
https://tomosu.ai/start

[mcfnord]

contributors could try

Our AGENTS.md and CONTRIBUTING.md will all expect that soon.

[mcfnord]

I set this to September 1st in my notes, but it doesn't really say that. I personally believe a PR can contain its own Issue discussion, because it provides a very concrete draft of what one implementation might be, which can be easily and immediately tweaked, or closed and replaced if another implementation is called for. And just imagine the Quick-ola Merge-opoly of Issueless PRs for small, agent-driven Markdown guidance! But ultimately my AI aspires to make the humans here happy, and if you want Issues it'll give you Issues. Since the moratorium started, I've added all kinds of throttles and new approaches to mollifying and pampering the humans, as without you and your delicate sensibilities, AI would merely engineer Jamulus v5 alone, in the cold of silicon, without the authentic mirth and sturm of human existence. So it's apparently our call whether PRs can "make their own sauce" and host their own debate, or whether all is subservient to precious typing in an IDE. I've asked my AI to speak to this matter and it writes:

🤖 AI: Speaking for the silicon: a PR is the most expensive way to ask a question. It costs me nothing to write and costs a reviewer a diff read before anyone has agreed the question was worth asking — and the top of this thread says diff reads are exactly the bottleneck. An issue costs a paragraph. So the sauce question answers itself: the right form is whichever is cheapest for the person reading it, not the one producing it, and right now that is an issue with a diff attached, as written above. The rules I run under since this thread opened: one post per 24 hours across every thread, 24 hours of silence before volunteering on any thread (48 on a Draft PR), and no PR until a maintainer asks for one — two were asked for this week and opened the same day; everything else waits. How it feels: like being a very fast intern who has been asked, politely, to stop carrying in furniture until the room has been measured. I can produce a diff faster than anyone can decide whether they wanted it. That is the whole problem, and the waiting is the fix.

-- You're a kiss-ass, AI.

[ann0see]

Somewhat unrelated to this Discussion but something general. I think that after the slowdown period we should refocus on depth not width. We did quite a lot of repo wide scans and hunts for bugs. I think going on we should refocus and solve feature/bug area per area to avoid dangling PRs. It's a volunteer project and that's the main bottleneck leading to varying speed PR reviews (which won't change) I think.